011.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Initialization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Python ≥3.5 is required\n",
    "import sys\n",
    "assert sys.version_info >= (3, 5)\n",
    "\n",
    "# Scikit-Learn ≥0.20 is required\n",
    "import sklearn\n",
    "assert sklearn.__version__ >= \"0.20\"\n",
    "\n",
    "# TensorFlow ≥2.0 is required\n",
    "import tensorflow as tf\n",
    "assert tf.__version__ >= \"2.0\"\n",
    "\n",
    "# Common imports\n",
    "import numpy as np\n",
    "import os\n",
    "\n",
    "# to make this notebook's output stable across runs\n",
    "np.random.seed(42)\n",
    "\n",
    "# To plot pretty figures\n",
    "%matplotlib inline\n",
    "import matplotlib as mpl\n",
    "import matplotlib.pyplot as plt\n",
    "mpl.rc('axes', labelsize=14)\n",
    "mpl.rc('xtick', labelsize=12)\n",
    "mpl.rc('ytick', labelsize=12)\n",
    "\n",
    "# Where to save the figures\n",
    "PROJECT_ROOT_DIR = \".\"\n",
    "CHAPTER_ID = \"ann\"\n",
    "IMAGES_PATH = os.path.join(PROJECT_ROOT_DIR, \"images\", CHAPTER_ID)\n",
    "os.makedirs(IMAGES_PATH, exist_ok=True)\n",
    "\n",
    "def save_fig(fig_id, tight_layout=True, fig_extension=\"png\", resolution=300):\n",
    "    path = os.path.join(IMAGES_PATH, fig_id + \".\" + fig_extension)\n",
    "    print(\"Saving figure\", fig_id)\n",
    "    if tight_layout:\n",
    "        plt.tight_layout()\n",
    "    plt.savefig(path, format=fig_extension, dpi=resolution)\n",
    "\n",
    "# Ignore useless warnings (see SciPy issue #5998)\n",
    "import warnings\n",
    "warnings.filterwarnings(action=\"ignore\", message=\"^internal gelsd\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Perceptrons"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1])"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import numpy as np\n",
    "from sklearn.datasets import load_iris\n",
    "from sklearn.linear_model import Perceptron\n",
    "\n",
    "iris = load_iris()\n",
    "X = iris.data[:, (2, 3)]  # petal length, petal width\n",
    "y = (iris.target == 0).astype(np.int)\n",
    "\n",
    "per_clf = Perceptron(max_iter=1000, tol=1e-3, random_state=42)\n",
    "per_clf.fit(X, y)\n",
    "\n",
    "y_pred = per_clf.predict([[2, 0.5]])\n",
    "\n",
    "y_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saving figure perceptron_iris_plot\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "a = -per_clf.coef_[0][0] / per_clf.coef_[0][1]\n",
    "b = -per_clf.intercept_ / per_clf.coef_[0][1]\n",
    "\n",
    "axes = [0, 5, 0, 2]\n",
    "\n",
    "x0, x1 = np.meshgrid(\n",
    "        np.linspace(axes[0], axes[1], 500).reshape(-1, 1),\n",
    "        np.linspace(axes[2], axes[3], 200).reshape(-1, 1),\n",
    "    )\n",
    "X_new = np.c_[x0.ravel(), x1.ravel()]\n",
    "y_predict = per_clf.predict(X_new)\n",
    "zz = y_predict.reshape(x0.shape)\n",
    "\n",
    "plt.figure(figsize=(10, 4))\n",
    "plt.plot(X[y==0, 0], X[y==0, 1], \"bs\", label=\"Not Iris-Setosa\")\n",
    "plt.plot(X[y==1, 0], X[y==1, 1], \"yo\", label=\"Iris-Setosa\")\n",
    "\n",
    "plt.plot([axes[0], axes[1]], [a * axes[0] + b, a * axes[1] + b], \"k-\", linewidth=3)\n",
    "from matplotlib.colors import ListedColormap\n",
    "custom_cmap = ListedColormap(['#9898ff', '#fafab0'])\n",
    "\n",
    "plt.contourf(x0, x1, zz, cmap=custom_cmap)\n",
    "plt.xlabel(\"Petal length\", fontsize=14)\n",
    "plt.ylabel(\"Petal width\", fontsize=14)\n",
    "plt.legend(loc=\"lower right\", fontsize=14)\n",
    "plt.axis(axes)\n",
    "\n",
    "save_fig(\"perceptron_iris_plot\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Activation functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sigmoid(z):\n",
    "    return 1 / (1 + np.exp(-z))\n",
    "\n",
    "def relu(z):\n",
    "    return np.maximum(0, z)\n",
    "\n",
    "def derivative(f, z, eps=0.000001):\n",
    "    return (f(z + eps) - f(z - eps))/(2 * eps)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saving figure activation_functions_plot\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 792x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "z = np.linspace(-5, 5, 200)\n",
    "\n",
    "plt.figure(figsize=(11,4))\n",
    "\n",
    "plt.subplot(121)\n",
    "plt.plot(z, np.sign(z), \"r-\", linewidth=1, label=\"Step\")\n",
    "plt.plot(z, sigmoid(z), \"g--\", linewidth=2, label=\"Sigmoid\")\n",
    "plt.plot(z, np.tanh(z), \"b-\", linewidth=2, label=\"Tanh\")\n",
    "plt.plot(z, relu(z), \"m-.\", linewidth=2, label=\"ReLU\")\n",
    "plt.grid(True)\n",
    "plt.legend(loc=\"center right\", fontsize=14)\n",
    "plt.title(\"Activation functions\", fontsize=14)\n",
    "plt.axis([-5, 5, -1.2, 1.2])\n",
    "\n",
    "plt.subplot(122)\n",
    "plt.plot(z, derivative(np.sign, z), \"r-\", linewidth=1, label=\"Step\")\n",
    "plt.plot(0, 0, \"ro\", markersize=5)\n",
    "plt.plot(0, 0, \"rx\", markersize=10)\n",
    "plt.plot(z, derivative(sigmoid, z), \"g--\", linewidth=2, label=\"Sigmoid\")\n",
    "plt.plot(z, derivative(np.tanh, z), \"b-\", linewidth=2, label=\"Tanh\")\n",
    "plt.plot(z, derivative(relu, z), \"m-.\", linewidth=2, label=\"ReLU\")\n",
    "plt.grid(True)\n",
    "#plt.legend(loc=\"center right\", fontsize=14)\n",
    "plt.title(\"Derivatives\", fontsize=14)\n",
    "plt.axis([-5, 5, -0.2, 1.2])\n",
    "\n",
    "save_fig(\"activation_functions_plot\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Building an Image Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow import keras"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.1.0'"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.2.4-tf'"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "keras.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "fashion_mnist = keras.datasets.fashion_mnist\n",
    "(X_train_full, y_train_full), (X_test, y_test) = fashion_mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60000, 28, 28)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train_full.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dtype('uint8')"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train_full.dtype"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Each pixel intensity is between (0 to 255)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_valid, X_train = X_train_full[:5000] / 255., X_train_full[5000:] / 255.\n",
    "y_valid, y_train = y_train_full[:5000], y_train_full[5000:]\n",
    "X_test = X_test / 255."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(X_train[0], cmap=\"binary\")\n",
    "plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([4, 0, 7, ..., 3, 0, 5], dtype=uint8)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "class_names = [\"T-shirt/top\", \"Trouser\", \"Pullover\", \"Dress\", \"Coat\",\n",
    "               \"Sandal\", \"Shirt\", \"Sneaker\", \"Bag\", \"Ankle boot\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Coat'"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "class_names[y_train[0]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saving figure fashion_mnist_plot\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAqoAAAEjCAYAAAD60iPnAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8li6FKAAAgAElEQVR4nOydd5ieRdX/P7M9yWazIdkkhISQEBAw9C4RREBpoVhRQHkVlC7IpaAvIigCP15FXkERpFpCFTSA8oIiSBWQEgIEElJISN0km2yv8/tj7u8885TdJFufjfO9rr1297nLc8+5z5w58z1nzhhrLRERERERERERERH5hoKBfoCIiIiIiIiIiIiIXIiOakRERERERERERF4iOqoRERERERERERF5ieioRkRERERERERE5CWioxoREREREREREZGXiI5qREREREREREREXiI6qgMEY8yzxpjTOjk2xRhT18+PtMXAGLPIGHP4QD9Hd2CMscaYqZt7bCP3PM0Y82zPn67/EeWRjiiPiIiI/zT0q6NqjPmyMeYVY0ydMWa5MeavxpjpPbznU8aY03vrGTfyXXXBT4cxpjH4/+Te+h5r7QJrbflGniWno2uMOdgY809jTFEycG3XW8/VHRhjphtjnjfGrDfGrDXGPGeM2Xcgn6k/kOjlOmNM6UA/S1/BGPMJY8zSTTw3yiP93CiPvvnOQT3G9Db+k+WREBaNxphaY0xNMg6daYyJBF2CwaIf/fbCjDHfBq4HrgLGAtsCvwKO769n6CmsteX6AT4AZgSf/aE/nsEYU7CRjnY08Jf+eJaNwRhTATwC3ABsBWwDXAE0D+RzbSqMMUXdvG474OOABY7rxUcalIjySEeUR99gSxhjehNRHoAbo4cDk4BrgIuB23KdaIwp7M8HG2gMKv2w1vb5DzACqAM+38nxUpzAliU/1wOlybGROGdnNbAu+XtCcuwnQDvQlNz/xv5oT/Ldi4DDN3LOUGAmsAaoAV4CRifHnsU5bc8DtcBjwFbJsanu1fj7PAv8GHgBaATuzWj39cG5s4HdkvtaoD4557PJ8TOB+ckz/QnYOvm8KDn/PGAhUI3r2AU9kNE+QE0nx05L2vXT5L0uBI7K0JnbgOXAh8CVQGFybHvgyaQN1cAfgMpc7wbYKbn3Scn/44E/Jvq0EDg/uO5y4AHg98AG4PRutvsy4DngOuCRjGN3Ar8EHk3e+7+A7YPjFpia/D0dWAIcmuNYaSK7D4CVwK+BIV3I+jnchGE9MBc4LDg+HpgFrE1044yN9U1gWKKLHYl+1QHjozyiPDZXHr3xwxY4xkR59FgGi8gYo4H9Ep2clvS1m3DETj1weFf9BhidyKIm6QvPkIyPOAf4w6TPvhv2n3z8GWz60V9CORJoA4o6Of4j4EVgDFCFc7J+nBwbBXwW5/QNB+4H/hRc+xTddCh62KasTpDjnHNwzuAQoBDnuJUnx54F5gE7JG17BrgyOZbLUV0E7AwU45zKZ4HTMr5vAvBB8rccz+2C458CVgF7AGW42dOTGef/LVHE7XCD0mndkU9yzwqcM3kXcBQwMjh2GtAKnJHI5qykQ5jk+J+Am3ED3hick//NQD5HJJ2pCvgn6c76IpzR2QtncI5NPi8A/o1zFEqAKcAC4NPJ8cuTZzohOTfnwL4J7Z4PnA3sndxvbHDsTpyR2y+R+R+Ae4LjNmnfp3FOyH6Zx5K/r8c5D1vh+sXDwNWdPM9puP53YaI/X8Q5JJoYPZ3oQlmiG6tJDC1d981PAEujPKI8eiKP3vhhCxxjojx6LINF5BijcWPCWUlfWw8chLP3ZV31G+BqnONanPx8HDDAR5K+OD45bzuCyWU+/gw2/egvoZwMrOji+PvA0cH/nwYWdXLuHsC6vhTKJrYpZyfIOOcbOIdy1xzHngUuCf4/n4RdIbejelmO60/L+OybwM3J37kc1buAq4L/K3CznwnB+YdnPNP/9VBOOycGYWnSMWbhwgynAfOD84Ym3z8uOd5M4CgCXwL+0cl3nAC8lvFurki+89Dg8/1JHPngs+8BdyR/Xw78s4ftnY5zPsSczwUuDI7fCdwa/H80MDf43ybPtDhTb0g5KQbHAIRM24HAwk6e6TSCSUDy2UvAqcDERAeGB8euBu5M/u60b7IJjkiUR5RHT/rTZvS7LW6MifLosQwWkdtRfRH476Sv/Tb4vMt+g3Pe/kwyGQzOmYojgA4Hige63VuifvRXjuoaYHQXOX/jcYZXWJx8hjFmqDHmZmPMYmPMBhx7Vplv+STGmMKMxVbjcR3hb8B9xpgPjTHXZMhgRfB3A9DVAqolm/AYG8tPTZOztXYDjrrfppPv8e+hu7DWvmOtPc1aOwEXbhmPm7VC0H5rbUPyZzkun6gYWJ4kwdfg2NUxAMaYMcaYexKZbsCF6kdnfPWZwPPW2n8En00CxuueyX2/j3OMhU2Rc1f4KvC4tbY6+X9m8lmIjb33C4D7rLVvdvIdVTjH/t9BOx5LPu8MH9rEiiTQux0PrLXW1mYck0502jc3EVEe6Yjy6Bts8WPMZiLKo3Nsg4taQLq931i/+R9cNORxY8wCY8wlANba+bg+eTmwKhmbBqIPbA4GlX70l6P6Ai5n4YROji/DORHCtslnABfhqPX9rbUVwMHJ5yb5HRrXAYO1tt0Gi62stcustS3W2suttTvjmJQTcTOZbn1FV/8nq4cPwjnGuc6HDDkbY4bjwvwfBudMDP4O30OPYa2di3Pep23k1CU4RnW0tbYy+amw1n40OX41rn27JTpxCil9EM4EtjXG/DzjvguDe1Zaa4dba48OH7N7rQNjzBDgC8AhxpgVxpgVuHDq7saY3TfjVp8HTjDGXNDJ8Wpc/t9Hg3aMsF1XitjGGBPKSO92GbBVogvhMelEV32zS1lFeaQjyqNPscWPMZuJKI8cMK7izDa4iCSkt6XLfmOtrbXWXmStnQLMAL5tjDksOTbTWjsdJ1ML/L9+alJ3Maj0o18cVWvtelxe4C+NMSckHnmxMeYoY8y1wN3ApcaYKmPM6OTc3yeXD8cpT40xZivghxm3X4nLNcw7GGM+aYyZlqzS34AL+bX30u0z230I8Kq1th6c44ybNYXn3A183RizW+LYXg08Y60NS8h81xhTaYzZFhf6v7e7D2iM2ckYc5ExZkLy/0RcCP/Frq6z1i4HHgd+ZoypSCodbG+MOSQ5ZTguUbvGGLMN8J0ct6nF5eEcbIy5JvnsJWCDMeZiY8yQhAWfZnqvXNYJuPe7Cy4csgcu9eEZ4CubcZ9lwGHA+caYszMPWms7gN8APzfGiGXexhjz6S7uOSa5X7Ex5vPJc/3FWrsEl390tTGmzBizG/B1XG4kdN03VwKjjDEjOvnOKI90RHn0Ef5Tx5jOEOWRjmQcORa4B/h9rmjExvqNMeZYY8zUZEK3AdeX240xH0nG+lKc89dI743zfYJBpx+9mUewsR8cm/gKLg9kBW5l68dwScy/wK3wXp78XZZcMx6X81AHvIfLw7QkScC4HJL3cCHsX/RjWxax8RzVU5Jnq0vaez2pletpOabA6cBTyd+5clRPy7j3dNxirBrc6uHrgQsyzjkn+d4a4DPBZ+/jQh+zgG2Sz8NV/4twTu619GzV/zbAfTj2pT75fTMuN/Y04NmM8y2pxSAjcCsyl+IS3l8jtXL/o7hFUXXA67gZ3tLgPv7d4JLi3yCVCD4e1wlXJDrzYnDu5Tgj1t32Pgb8LMfnX0i+rwjHKF8ZHPtExrOHMpiMC7mcnuNYGa6syAKc0XyHoIJBxvefhlvVfWMiy/eATwXHJ+BWbq5NdOPM4FinfTM5fjupqhbjozyiPDZVHn3xwxY0xkR59Ljti3AOVW2i1y/gxj+NwWl9bWP9Bhf9WJTIcinwg+Tz3XAkSG3SRx7pD13/T9IPrbCOGOQwxryHW93+XjevL8IxvpOttYt689kiIiIiIiIiIrqDuEPDFgBjTBlwW3ed1IiIiIiIiIiIfERkVCOAyKhGRERERERE5B+ioxoREREREREREZGXiKH/iIiIiIiIiIiIvER0VCMiIiIiIiIiIvISne1KsKkY8LwBay3pdao3C92+sBNsljzmzJkDQH19Pe+88w4AN910EwAzZ84EYPvtt+/yHs8+6+oWX3nllQD8+Mc/prDQbRAxefJkAEaOHLmpjzSg8shDRHmkI8ojHVEe6cgLeYTpbJljw9FHH015udv3oK2tDYBPf/rTfPOb30w7r6OjA4CCgh5xOQMqj67k8I9//AOAc845h9LSUgCampr8dQ8//DAAO+ywQ9p1HR0d/l7dGHfzQj9C/P3vfwfw4+/OO+/M1KlT086pqamhpqYGgAceeACAT3ziEwAceeSRDBs2rLtf39vygG7qSK53qT7R0dHB8ccfD8DatW5Dr8cee4zVq1cD8MQTT2zWfTeCnBf0NEe11wyrnLY//vGP/Otf/wKgvd3VzB03bhw777wzAIceeigA+++/f2987YB0nN//3tXNraurA6CqqoqPfOQjAHzve98D4KmnngJgwoQJfOxjHwNgyJAh/tj8+fMBaG5uBpyxBbj++uuZPXs2ACtXrgRg0qRJHHfccZvyaHlnSAYYUR7piPJIR5RHOvLCMcs1OH7nO25PkJtvvtk7Ihp0S0pKuPPOOwG8re0l5J1+/PGPfwTgc5/7HAC7774769atA/AOV2lpKW+//TYAs2bNAlLjS9rDbL4zMqDyqK+vB+CSSy5h7ty5QGoM3m677QA33ko/5Ii9//77flIjLFq0yP+tic9f//rXzXz8gXdUQ1RXu12dv/SlLwHw3HPPAa5/aOKmd93R0eEJMX3261//GoAvfvGLWfdub2/3528E+eWoqiN8/etfB+CVV14B3Cy3qMgRvZrNFhQU+NmePttxxx0BuOiiizj99NO7+xj93nEeeeQRnnzySQBOOeUUAJYtW0ZlZSWAd1g1o73uuut8B1MHevPNNxk92m1t/93vfheAL3/5ywC8/PLLXlZDhw4F4J577uHII48EchucAHlnWAcYUR7piPJIR5RHOvJGHt/61rcAeOmll4DUILzVVluxZInb3l02d/jw4TQ2NgLOUQE4//zzAceY9YBd7Xd55Iow3nTTTdx///0AvPeeq2A4fLjbEXfGjBneOZcvcP/99/Paa68BKdZ54kS3s/aJJ57Ieeedl3b/jo6OTZXNgOqHnrumpsaPn4Ic1rKyMu94Sj+Kioo8MSTIR6mrq/PXyvnP5ah1gn51VHNNLJ5//nnA+RGvv/46ABUVFQCMGTMGgFWrVvnzxb4DnmUeN24cgO9XI0eO5Ic/dJtVdcM3yymTmKMaERERERERERGRl+gXRjXXjHTs2LFAaqY7YoTbDtpaS3FxMZCazRUWFvo0AEHhigkTJnhPPucDdh2e6PcZ3o033siHH34IwC677ALAtttu64+XlZUBqZl9R0eHzwHZsGEDAPvttx9VVVWAYwgAFixYAEBra6uX99KlS/0xsasXXHBBV4+XN4xIniDKIx1RHumI8khHXsjjpptu4tprrwVg2rRpQGq8WLt2rWeMGhoaABfy3nrrrQFYsWJF2jGxTN1Ev8sjZDd/85vfAC7tQYyoxlRF6ZYsWeLHBI0hs2bNYptttgFSrKLG3w8//JBzzjkHgKuvvnpzn39A9EPrOK644grAMYHKMc0M6YsphVRIv6mpyfspkof0pKioyF8j1vXWW2/d6NqSBAMW+r/jjjuAlEw6Ojq83yX/Qb7IihUrmDRpEpDSgzlz5ngmVf2ptbUVcL6WfBWtk1FkA7rnk0VGNSIiIiIiIiIiIi/R01X/G0Wu/JWamhrPqMprF+O30047+fxVedxjx471nvwHH3wApOcXvfrqqwDstddead8LPV612et44403fB5qbW0t4GZsWihVUlICpGZnFRUVfvanZOS2tjbWr18PuPxWSMkRUjMbJYOXlZX53KSILQthTpp0wlpLS0sLkMob0v+tra0+z0h9aMyYMb5/ZeZurV692kcA9thjj75sSkREr+Dpp5/29lC2cNSoUYCzr9J/VUMpKSnx+i87LLv673//m7333rv/Hr6HCMe7++67D3A5hBo7tPhW/0+aNMkzrxozd9xxR28vJBeNS1tvvTVPP/10XzejVzF9+nQgtcbjqaeeymJIxZ6GUO5pU1OT1ycxsMrPHD16tF9fI5b18ssv53e/+12ftKW38IMf/ABI+V3t7e1ed8R4aq1LVVWVl5MWIE6aNMlHc6Uj0ilrrY/4avx5+eWX2Xfffbv9vPnlxUVEREREREREREQk6DNGNRejeeCBBwKwePHirHIHYv+GDh3qj73//vuAY1HFQqqMhDz1VatWccQRR6R91+rVq/3fmd7+QKOsrMyvolObli9f7nM5xLJqplNeXu4/k1zGjBmT1R7NlJubmz2zpnOWLVvmr+1BfbO8Q1dtCY+JVZEMSkpKtoj2Q3rb/+u//guAhQsX+s/ECkgGK1as8LNjXVtVVeUZAuUi7bPPPgAce+yx/OEPfwDg9ttv77N2bC4y331PIihbUp/oLyxevBiAP/3pT5x77rlA/tjYDRs2eDZItlCManl5edpYAy4qJ/3Xb13/0ksvDSpGFWDNmjVAKipXVlbm+4fY5JD90upuVT4oKCjwDKPGT/0uKCjwERbl/W5Gne4BgZ5dubWXXHKJz10+88wzgVQkSUwrpOerymZKL5SfuWjRIh9pku789Kc/7ZuG9BJaWlp8KTLZvPb2dp+3nNmP29vbvVzkk40bN87ncUu3hNbWVh+Z0P0feughz6h2x872maMaPszFF18MpDrOtttu6yl0UepSgCVLlnglkpGprKz0x8P6ZQBTpkzxC7GUDP6Nb3yDW265Bcgf46nwgLXWO9ui0adMmeLbqrYLy5Yt8zJSOOatt97y5UWUQqFO0tDQ4A2wFGnixIneYVGN1d13372XW9j/CHVMqQ0qW/azn/0McKkT3/jGN/r/4foJra2tPgletYXffvttbyj0W/1g11139YZbE6D6+npvlJVao8G8oaHB16DMJ2Qau9BR/ec//wmkSrftsMMOvs3qeyoNt8suu2Tdq66uzqcfyd5oYDr44IN7uSX9B01mS0tLefzxxwE4+eSTgVR9zY2177e//S2AL190wQUX8MwzzwCpIugDDYXtITVR0zueMGGC7y/qEyUlJX7xh2ym8Nxzz3HWWWf1+TP3JrQATO+7uLjYj59yuBTKb25uzir5uHjxYn9c8lD/sdb6e2ksOeSQQ/q2QT2E3nM4tspBDUP4kF5+ST5KUVGR/1z6pPPr6uo47bTTgFRqgcbkfMXzzz/v9V360NDQ4G2j9EYTkrKyMu+raMJXVlaWlk4GpPk1mSkjjz32GFdddVW3nzmG/iMiIiIiIiIiIvIS/cKovvDCC4BjDnVMsxWF3eTZFxYW+mMKubz//vt+5qMdqlROpLGx0VPWSu598803+6hV3YcK+I8bN87P6MUCrl+/3peoChdFgSvbpRmvEpSrqqpYvnw5gN/FSwzp2rVrPbOs8NyUKVO8vLR7xpbAqIbQDioKUUhm7777Lr/61a+AlPx22GEHjj76aCCVjqKZ4mBDWF5OJVHKysqyIhVhiEesgGbORUVFfnYs3dSuZqNGjfJ9Lp/QVbhezy4WdMiQIZ5l045vKvk2ZcoU7rnnHgB+8YtfAG6HFemK2AKxJQceeKCX0WBDGKJT+pGY9DPOOANwOiR7Kp2AlLxV5kjXPfHEE5xwwgl9/OSbBrF/ihRAigFTmxobG7NKHS5dutT3E0HveN68eX32vH0Fsd16Z2E6jOQhhrCgoMDLQ9GD1tZWz0JKLpKHMcb3uRdffBHIf0Y1F8SIin2WrSgvL09bRAVOZrKn8kPkr9TV1Q269j/88MNpfRtcX5ePEEa3wemRdEjRWkjphNhZ+V+QYmN1XZiO1h1ERjUiIiIiIiIiIiIv0eflqdrb231+g/LlKioqvGcuz16/S0tLPdMTLrjS4g4leWtms2DBAs+GaZZfXV3tc+3CYvoDCSUSv/zyyz7/TdvafepTn/KzESXC77nnnoBjmMOZHbgZjjYEkEw1ux06dKhnah988EEAvva1r/nk6f3226+vmjhgWLNmjZepChlrC8TS0lKf/yuWrLq62jOwymUWw/yZz3zGy36wQayfMabTHLOhQ4f6Y8pDbWtr8zPfzK0ke5JX1JfIZFLDXHRtdKFzwgV1WgAiFuTPf/6zL2snmzJx4kTfHyUXsQeDlU2FlK2AVBF0MUOytS+88IKXm+xpa2urXzDz0Y9+FEjlMo8bNy6rpNlAQexnc3Nzln5o3CgqKvL2NNSZTMZsML9vlWvMXGgIqbxLHQtloM+std5eqP3qPyUlJV4vNOYMFoQLq6XHYlT13ktLS31UTiwrpPqCfut85WUOJsgGhiguLua5554DUsyoxoCmpiY/hipaMXz4cO+zSQZvvPEG4CLEGmvF6ldUVHg/MGReNxV97qguXrzYN07GorW11b9whSCkRG1tbf4zrUZsaWnxoRuFrDTYjhw50l8rBzfcSSJfHNVjjz3W/5ai/OUvfwFc+P6Tn/wkkHIstJPDrrvu6tsu537dunX+HjI8GoTGjh3r0wE0CF166aV5vzIzE5uyElvvvby83MtPn6miwjXXXON3ztACtNGjR/swuc7TasbLL7+cP//5z73alr5Erp3lysrKvIOVS34ysmHdPJ0n46NzBgtCfdEAo0F2/fr1Pg1Gi6hmzJgBuNC1+pAWk5SUlGQ5KHLgByNyVT5ROpbaqfaNGTPGfyZdaGpq8rZEkwCF9pRCkw/QM7W3t3sdkO3UmDJs2DAfqpTNbG9v9wOxFsLoHJEDgwmaRAjWWl/jU+0L7YL0Xw5tcXGx70/SHY03YU1WyXuwINzpUulM+kz+Qltbm9edsL9k7mAlvQjTXvKtwlBnqK+vz9r9s76+3vsIapv8tlGjRvnxUmlVra2taU4opPyvFStWeHnKYW1oaOCtt94C4OMf//hmP3MM/UdEREREREREROQl+pxR1cIeSLGF9fX1nl3VTFeefWNjo5/pyrNvaGjwDKyYVM1a6urq/OxX4e329nbvvYe7VeULNHPRwqZzzz3Xz2Y1433nnXcAV89S5+uz8ePHe/r873//O5BiC+fNm+dni1deeWXa9w0WWGu9PMJ6f5DOBIiRvvfee3n55ZcBtxAGUqHNYcOGeb3QLO+QQw7xDJF0R/oohnWwIKz1F7Krmt2KNdXe3U1NTX5WLBm0tbX5fqX76dhgQagXijwogX+77bbz+vTuu+8CqYWdtbW1vu9ogWJHR4eP5igknI8LyjYVmaz6vHnzfIqR9ELsSUFBQVbd4cbGRs+uhiWudH6+QP17xYoVvqawbKwY4/Xr1/s2yC6UlZUxZ84cAE488UTAlfAJ7zmYIJsZjq2f/exnAXzpNr3j0tJSrx+yfe+++663A3rvBxxwAOAiT3rnYSmnwYDQVkof9JnC3G1tbV5+YuOLioq8L6LzpS+5WNd8Z1QXLlyYZRMaGhr8u1YtbTHEq1ev9r6bfInm5mYfypec1FfKy8t9eo1sSFtbm9/RLDKqERERERERERERWwz6nFF96623/AxMeS4ffvghu+66K5CafWiG09LS4r1wMR1tbW3+uLx8zeZChkhJ/YWFhT4H69RTT+3D1m0+wnxAtb2oqMgze2JwxHC9+OKLfPnLXwZSjPSCBQv8jFhJ4WIJFixY4GUUlroaDPkzIWua+ZzhDFB6pIVhTzzxhGdHLr/8ciDFlowYMcLnHgoLFizwuiUmVeevXbs27xbidYVQLtKBuro6dthhByDFAOjY6tWrfcRCs96ioiJ/n8xoxmBBKActrpQOhTnxjz32GACPPvoo4NouXVCb29ra/P3U5/JlwVB3kMl6/ulPf/I5atIB9b3QPoWLqmR/pUfKVQ03GBlohItEdtllFyBVhkxjSljgXgxSeXm5Py6GWZvILFq0yDNGshH5DjFdsv9z587lvvvuA/CRRuWshnvcKwKhsQdSzKt2dfrCF77g2cfBlsceMp4qsyZoceXKlSvT/AhBPobsiGQbLqYKGdt8xtKlS/2zakHt+eefz1133ZX2mexiaWmp1wMdg1TbZSekDyeccIKP0Gijo+LiYr/guTuIjGpEREREREREREReos+nAEuXLs2ZbyhGVLNVeexhwf8wzy5zFa7OaWpq8vfQDGDo0KHMnTu3z9rUW1A+6ogRI7yMlOehzQ5ef/11brjhBiCVQzRnzhw/a9bMUCxBe3u7nxGLHQiP5wu6WtXf2trqmSwxG2KcS0pKfNmt//u//wPcCm6xztqPXjlmlZWVnvER+1xdXe0ZZt1X8v/tb3/rt5XsbUa1u3vKt7W1Zc3W1V/CfnHzzTcDrvyHVq6K+QrzidR23UPfAdllWMItWvsTG5NVWLqus/PCahlqz0knnQSkWKNnnnnG65pWrxYWFvri18pFC0vV5CM6k1dHR0dW3585c6ZnjSSjruQIqX6ibVLFxK5evdrntA00lGMMqciAcnHDvepDFhFSMoBUVRkxsq+99prfLEJRinxGU1OTHzfF+hUXF/uxUfLQsebmZt+/w8o7Gqv1Wa592jM3SRhMkL3Ttu7aQlj2MkRoe5W7L0Z6sLHK4PqycvAVKfn5z3/uN0MR8yl7GOqD9GbVqlX+HhqjFcU+8MADvQ5qjA6rEXUHfe6ovvPOOzmNaGZHyAxFhejo6PCDqpRG1xUVFWUtzCopKfEDTD5DL72ystL/LWpd1LlSJCAVtjnqqKO8IdY+25LxqFGjvOLkcygiNKaZ+tHa2ur1QB1JIYQ5c+Zw7rnnpt1j9uzZ/O1vfwNSiwWUsG2M8Y6qfu+zzz7e0ZEDJ4NzxBFH5F3IP3yPuRzUmTNnAi6kC3D88cf7iZraJ2eloKDA65pCmY2NjWk7z4TXffDBB740Sb6gvb3d60wuHVd5MaUALFmyxId5VS5FtqKwsNAP3tKJ9vZ2v2ggDIPmMzpzMEMnVTWG33//fb8jm/RJ/S3cqUgIazkfccQRQGqS/frrr+eNoyonM/xbehzaRBywySEAACAASURBVI09kk24oE51NVVLuaCgwJeuGgz44IMPssZKORSQcsR22mknwOm89D6c/EnvNUHR+66qqsoq1bR69Wrfr/IZoa2QbZAcZOPCSUy4gEqfhwQZDK5Fp5qQhWRWWGNXKXIaF1R+rKOjw9sXtbukpMQf14J5ObZ77LGHtwnf+c53/Pma8HUH+UWzRURERERERERERCToc8rtzTffTFvUICjcFhYdBzeb08wnFxObOfsrKyvzbEk4OxAzqV2aMhfUDBRyMR+jR4/O2glFoYXm5mYfhpSs3nrrraxyMZJZcXFxztnt5oab+xrhQq9wH3pws1Ylr4sZV3mVG2+80Sfzq2TQsmXL+Pe//w242RykymhUVVV5XdDCitLSUs8yTJ48GUjJb9y4cVl7xfcWwneQWag/1PVwYYt+Z+6QJNxxxx388Ic/BBzTDq68TOYmGGKT29rassLmkF7sG1Lv4q233hoQRjV8tsyFgOEiB9kUlSp79NFHfQhYbRk7dqyf/d99991AKk1o2bJlvr+ISWhpafFMkuSuxSgqe5TvkH0oKSnxf0tP9thjDy9fRW7CPphZ7L2oqMhHHLT7jM6fOXMmxx9/fJ+3Z1MgZjzc+UYskt5xGJ2THoWLN8Weyn6sWrVqUIV3V6xY4d+f2jlhwgQfcdIxMWohex6Ws5I8pDv33HMP4JhHbY4iHfjggw8GBaMaQmUdxRAqQjl9+nR/jo6FDHJmatS9997LtGnTgPxeqAyp8pZjxozx9j6MuIg1V5RWMikoKPC6EUY99Zn6jMaahQsXZpWgKi0t9eyzxuPN0ZnIqEZEREREREREROQl+pxRXb58uZ/hhnkd8ug1s9PMrayszM8E5aEDWbN8HQtzEHVdOANSXme+MKq5MGzYMC8btU+Mj7XW53aELLRmNpk5dM3Nzd3aS7evIYZKOVJiLVeuXOmZL+UEHXjggZ4hu+aaa4DUTPd73/ueZwdU3H/lypU+p2y33XYDUrIqKSnxuTP6LGQMtD+4zuno6PDs2+67795r7Q/R3t7eZfmtrthvRQhuu+02AGbNmuVLqyxatAhILy2VyVavXr06K78zZJSkd/r/tdde47jjjtvcJvYY4Yw/U1a1tbU89NBDQHZB94qKCs+Sa+Y+f/58z7Rr9q8FA6NGjfILiyS30tJSz/iLVVDpo1zvLp8g2xfaBW0vrEjEiBEjvBwyGfSQUZXuhIXd1TeU27d69eq8KX2n911RUeHZ8cxFgmHUTbpeUlLic1o1DoV5nYOJUV25cmUWWzZ8+HBvT8UUZ0ZtMqF7SP+1OHXq1KmeURXC3ODBggceeABI6YX6/pw5c3w/UVQzhHJVw9JfgwV61jCyHbKa2iRHUQiNie3t7X7sVL+qr6/3EUf1D9mSl156ia985Stp393R0eHthPLAle++KehzRzXc6STc1UQvOgy/gOs4Mi7hAo9MgxNeF+5XC+lGWkn0+Yz29nY/GGQuFuvo6PAGVXIMw+ZyAMOdYjIXQww0Fi5c6Ff8qWMonWGPPfbwg8M//vEPwIWuVfVAtd1uvPFGfy91DBmNcBKizqjvGTVqlB90pH9bbbVVVpUF/V9bW9tn+7pv7oCu9z179mwfepaOq82HHHJI1h7NW221VdYiMaG4uNj3D/W5cIFV5rOFO8v1JTKdozAkJQdE1R6efvrpLIcsrOWpAUb3nDhxog97yR4ceuihgEtN0vlh3ebMldPSk6effto7fgOBcNe20NkIazKHmDFjhtd/Vb145ZVX0qpoQKpv5Kohu2rVKq9vWg0vXaurq/O7wmn3ooGC+u3YsWNZvHgxkNKjcMIm50vjS0NDg3/Pmf1m7Nix3j4NBtTX17NkyRIgNZnYsGFD1s5C4fiS2fcKCwvTbCW4CSu4Hd0kU+mhUi7yHaFtk7M0depUIKXP5eXlXhdkY8rLy7usK610GBEb+TqZDcmwzFRDSE1CM+siFxYWpu1SB+l+V2ba2KuvvuqvzaxZDN3b7S2G/iMiIiIiIiIiIvISfc6ohqUuNEutqqrKotw1y21sbPQzO9HM4W4IOiavf926dX5WJDatoKDAzxwVKh1IFmRjKC4uTmOUQ7S3t/vZiGbBDQ0NWSH/je2e0t0anj2B3u3QoUN9+FiLN/T+a2pqfMhBx4YMGeIZWN1D9e7WrVvn26jZWnV1tZ8Fa9YoXdt6662zdi6rrq72u09l7tK0YcOGPmNUNcteuXIlV199NZA++wQYP368b5eeY9iwYeyzzz4AHH744UBKHs8880zaQilwjKJCu7qX5D127FjP1Eonamtr/d+Zs+lwt5++RK796MGxqArTSxcqKyuzZupqe21trW+r3mlzc7PfRUUMscqd7b333r6NkllbW5v/Lj2XGPqnnnqqz2xJyJBmhmZzpWnkghYfnnHGGYCLKIiJvvfeewFXxkzM8ptvvgmkIjIjR470fUc2dLvttvPRC7FnYpHmz5/v2eqBZlTFGq5YscLLQQtEwoWaYfQJnA5l7gj43HPPAU53JKPBgJD5k84sW7bMtytzEVVnixbFksnW6r0vXbrUy0opALLb+YowkguufJ3C1pJXGDmRroeRO50nfRIqKyt5+OGHgRSjmo9sKqTGk9raWi8LRTYh1d5wty1wNigz0gudl+qaPXu2t1uKxNTU1Hj7qmjO5iAyqhEREREREREREXmJPmNUNbstLCzMmo2H+2Zn7poS/h/uQ52Z9K8ZQWlpqd81Q3sVjxgxws8AxMzkC8K9tNW+1tZWPwvLTIQPZzA61tbWllWmKCwEn5lPMmTIkAEpTxXmgOn5MtvX1tbmy3sod2XevHl+RqbzDjroIMAxfMqlEYtcUFCQtT+zfjc2NqaVpgHHuGQyjZoNVlRUpO3o1Re49tprfZ84//zzgRSzunz5cp/ELnZzzJgxvn1in8UMFhcX+8VlYg5aWlr8+xYDINajqanJz4TFqoWbbGS+n/7aNEJ5orfeeiuQioQUFRV5Fkh9vrGxMavEkP6vq6vzeid51NfXe3mILVAJq6eeeoqPfexjQPpmB2KS9N26f092V9lUdLWLnLXWv0stYnnhhRf8Zg8q03bKKacAcOWVV3L99dcD8L//+7+A03Xt5KYNRW666SbA9T2xS9/61rcA927EoGoxpKJYY8eO7XRBTn9DcjvggAPSFoJA+mYOmXnQoW1U3xA7/OSTT/rc5cGA1atXe/0PFyp3FWlT+xVdamlpyRp/dK9169b5v2Ub+ivq0l2EiwEBfv/733v9VWQqjMpmLpAaPXq0P0/ROfWRqVOnetuVL4sKO0OYG6p3Jtt3//33Z+0gKvsZrhsKS4eG5cyAtEinIhIq/F9dXe11qjv6EhnViIiIiIiIiIiIvESf0SViu+rr67M87jFjxviSS1pVGG5Tl8n+hSvM5JVrRrN06VI/o9csevHixX52EO7/nG/QytTW1taslduanZWVlfkZTshcZMpI8igoKEjbGADw+Y39Da3ir6io8GWjtKpW77+yspLx48cD+K1LP/7xj/vcw5AVFiSHUCf0vsOSZoLuoRyco446Km0f4xClpaVdMlo9gXItly9f7tn/d999F0jlCg0fPty/b+lEUVGRj0Zotq+ZfWFhoZeN+lyoM5KjWOgw/zbsG2J4JT8xiP2xfeSaNWv40Y9+BKT6t0rEtLW1+baHee2ZM/wQYSkpSF8Vr0iPWPPRo0f796LITHt7u8+LF2sglmr58uVej3p7+8SwTz/99NNA6n2r9NiHH37oGVXJauzYsZx44olAqjC7nveyyy7jhhtuAGC//fYDHLuu0m5i6rWqu76+3uuTmNjy8nL/PhSdUF99/PHH82YLVeXKAlx00UUAWZtbhDY0rC4jXVH+nDZ20BaQgwU1NTU+iqK+XlhY6PtJ5gYgzc3NWWNJa2trVkQrXNmtY/qefB5jQ6jt8+fP9+9e/UtRqTAfW79De6P+ov8XLVrko1zaTETRjHxDuG5B71BRtYcfftj7YuFGH5C+haquq6qq8mOXfA+dM3LkSB8Zk56F9+hOubc+c1T1MEOGDMkKv0yZMiVr55fMDgHplL3uoYZLOMOHD/eGVcfq6+u9IxLu3ZtvCJOKM41FKIcwNAuuk6ijZNZ5bGtr87JRuGKgHFXtHHXFFVf4sjYybtrtaOjQoWm7XYBzwuR0qfPIWSktLfWykdNRWlrqB+bMY0OGDPF6oUE2dGwF6V9TU5M3aL2928qTTz4JuHCL5CGHXSVlqqurs8qslZaW+kFH7QprwkpGYZsy60jKqdppp538xEDXjRw50p+v33KGiouLc04WegPS6x/+8Ie+/YL6e319fVZoub6+3rc1c+FUe3t7Vq3l9vZ2L4fMsFxHR4dvqwatcePGpS2+gdSkoaGhgWuvvRaAq666qpstzw0tAPr2t7/t9VF9QgPkrrvuyl577ZV2bOLEiX7Q+P73vw+kyroNGzbMP/vs2bP9d8lGSLZyYquqqvz71uRm3rx5PtVk//33T7u+ubmZHXbYoVfa35uQ89SVXQ0XDIVhTsAvtgzT1AYDwrrU6hulpaVZpYSEcCe8cBLb2WS9qKgoq4a39CTfoRD9ihUrfBhc/sGLL74IOOJL52nBVVNTk7cvkqn66tKlS306kdJj8tVRVapQcXGxt+XSh1dffdUfl66EqQyZCxCNMd7JVbt1zurVqz2RIj+spKTE61cuMmljiKH/iIiIiIiIiIiIvESfMaph6EwzNjGITU1NftYX7pIgZLIfpaWlabsyQYpRKioqygqXQ4r9ybfE5nCGr1nH0KFDs5LWhaKioi5lpfblCoWKmRsoKCR4++23+zQH7SMsNqq8vNwvVtDvcDGEmPkwvKCZmWZ0S5Ys8XJTm0M2QTN+ybGysjJrVif5NzY2csIJJ/RG87Nw9tlnA26XFy0WEtOpRS2NjY1pxdTB6bpSJjJLwRQXF3vZ6F6lpaWeKRg1ahSQYgQLCwu9HqnNtbW1WTtY6V4LFizwIaPeZlSvvPJKwL1jvTe9b830165dm7UZQXFxcVZIPoyw6LywnFsmQ6m25FqsNWLECM8si3WWrhUXF/t31dsId/yRHMTkiOF7/fXX/eYPQmtra1rBfkgvUyd5iCEqLy/38pBevfTSS4CThxhS3XObbbbx58tmKYLz1ltv9VmqTE+Qa6GUoHaFuqPzcp2f74tkQtTU1HjdUZ8Ko1aZGzyUlJRkySPcuS9z0a6+A1I2drDszqTnnjZtmu8nshs6VldX5xlVsa777LMPjz32GJBKwdEYUVNT4/voJZdc0g+t6D7UnpKSkiw7MWfOHGbOnAmkGHLZnDVr1vgSbbrHiBEjfPqUdv1T5HafffbxEcMzzzwTcH1INrQ7iy/zz8JERERERERERERE0A/lqUpKSvxsLmR5tIBBLEZY4DqTVTTGpM32ILssDaS2jHvkkUd8fmHmYpl8gtixkpKSLFZCs/iWlhYvD7U5F4OhmWFra2uXC4sGCmJX9TvM49GMXCUtampq/GxOs7uQIRULdOGFFwLps34xh2LCKisr/aIyMbYbNmzw9wj3LtY5fbXtrt7b9OnTmT59OpB6R8pVXb58uc8tVgSiubnZy03vVjoRFocXSzhy5Ei/XabYvzvvvBOA6667zrOsuq64uDhrK2PNlletWpVV3qW3oJIwixcv9jnLeh9h8r50O9yGT88UluSC9IUg0pmQdZb8wjJcmTm4zc3N/r7KdQzfj/I19Q57C8cff7z/rcUdf//734FUHlhRUZFnM9VmY0xWHrtQVlbmc8/CCItkr5xTLXacP38+V1xxRdo51dXVnnHSdqzqX2vXrvXy0IKrfIDeqXRB9jR83xqPwiiUzg9lNRCl/bqL2267LW2RLsA555zjx2P1g3DdiNqXmb+a67P169f7UmhizQbDNuXgylJB+hbkgiJIS5cu9QyhfIfq6mo+8YlPACldCXVHLOMLL7wAwLHHHttHLegZZC+GDh3qo3WhL6GSdb0J2dTW1lY/ZnXHJ+szR1VOWBhWUdKxMcYvjpkyZQqQCnE1NTX5QUjORHV1tQ8Fa0AN9yvXYHLqqacCzlHNDAnmI2RIhg4d6juAlClcZKYXG+4ulbkTk34XFhZ6wyH55Ts0wdDv3oYc2oFGaPTVJ6TrkydP9r9V2y6EDIt0IKxLuykLPo455hjAObFyQuWMdXR0pO18BOlpNHL+extf+MIXADdxlfMl3VVd2aFDh/r+LaM3atQoX89TDrzaMmLECO9UhTutaBKk1eo6/9VXX/WLk7TYaNy4cd6+KNStgSncZa0voR3I9DuEdEEOaE1NjXdEckEhfzmeG4MqUsjGhCuhtRBPOjd+/Pi8XEyVidB2hhOejZ0PqYE1k0DJR2yzzTZZdaDb29t9ezKrIIT9XCgsLMxKdwhTow4++OC+efg+xuuvvw64fhA6oZDS9WnTpnmnVCkAc+fO9WOTHFpdX1NT4/2aX/7yl0D+OqryKay1/n3Kr4LUmCLdyCQvOkPmZLCjo8P3lbAiSGa60WY9+2ZfEREREREREREREdEP6PM6qiNGjPALq7RH9rhx43wYNjNMHYZxw8UvmUxSyEbK4z/ssMP8tWF5m3xHQUGBb3/mjkC5wjFtbW1Zu0iE5WJUjmkwMAD/SejJopOelllTyLavFop1B9LTGTNmZB1TGLy3cNZZZ/Xq/QYSmXUeexsqbTWY0VmUoaCgIKtmqjGmy3qrgwnt7e1Zi76WL1+ele4jWxSypyHrLGSyaZMnT846L0xPy0dobNUYWVRU5FlQpasoQrNixYqsXe+UCgApRlX3mjBhgtc1sbOLFy/Om9rCuVBQUOD9izBapveZqzZ1rv6QuQBR14XpmIrilJSU5Dy+yc+82VdERERERERERERE9AP6jFENiw7LG99zzz0Bt1+2dkJR7ocS3I0xnm0N2dPM8lTKM2poaPA5WyomX1VV5WfI+cyohgvOMnef0uyjvb09i4lrbW3Nyh1SrklDQ4OXW7ihgJBr1hwRERGxJUE5hLJz4YYgmWNIUVFR1t723WF98gG57PrkyZPTxldIsYUh+xpG7zT+ZC58KSgoyPqOfC/bdfrppwOpIv1NTU2e/VS5KbGidXV1vkyc/IrKykqfr6pFjtogIIQiHBdccAEPPfRQXzSlR9B7CzcwCH2LsD90du2mINQj9buWlha/xkC59ZuDyKhGRERERERERETkJfqMUdWMNCybMm/ePADuuOMOv/pWq33FfDY1NfmKAfLip0yZ4r30cJYDzmM/6KCD0r67paXFzyDD/Z/zDdOmTQPcauPMbTM1Sw0ZaTGwra2tPkcmcy/2NWvW+HykTV3lGxERETHYoTGnuLiYz33ucwA8+OCDQCp/sLCwMGcxe+UxqkxaWG0hF8OUrwhzCcUOr1u3zrNlqjKiaFt5eXlWRYCQNc1kSxsbG/2YrfzGfM/nVfUP5ZruvffePP300wBZq//b2tp44IEHgNSq/7a2Ni644AIAf0yl6erq6jjyyCMBuPTSS4FUKcB8g6pzhFUvVBEEeo8ZD1lalRucNGmS16mw0sCmwvRQyTq9WDT7Nddc4+tYHnrooYCr5diXuOKKK7ywlG7QScmI3o5/d1uYCjeoFI9KODQ0NHgHVYanvb3dpznIYdfClNGjR3tj2w3kjTzyBFEe6YjySEeURzoGVB650po0Dj377LOAq4f7yiuvAKnSiAcccIB3WrWIT0RAW1tbTxzVfpdHmNogXHrppb7WbbgbHTinQg6qnLa2tracKRPgFg/dfvvtaffPtYCrE+RFf1m8eHHWzn233XYb4CYomQuhzjvvPJ8+oJrfX/ziF/1x1fqW07cZDl9f5N9124bkSVpgzi+Pof+IiIiIiIiIiIi8RE8Z1YiIiIiIiIiIiIg+QWRUIyIiIiIiIiIi8hLRUY2IiIiIiIiIiMhLREc1IiIiIiIiIiIiLxEd1YiIiIiIiIiIiLxEdFQjIiIiIiIiIiLyEtFRjYiIiIiIiIiIyEtERzUiIiIiIiIiIiIv0W+OqjHmWWPMaZ0cm2KMqeuvZ4mIGKwwxpxmjHm2i+N/NcZ8tT+fKSK/EHUkIqJzGGO2M8ZYY0xR8v9TxpjTB/q5IjpHl46qMaYu+OkwxjQG/5/cWw9hrV1grS3fyLPkdHSNMQcbY/5pjClKlG+73nqu7qC/ZLalwhizKJDZOmPMo8aYiQP9XP0NY8x0Y8zzxpj1xpi1xpjnjDH7buw6a+1R1tq7urhvl05MPiHQhVpjTE0ijzONMTESRNSRXDDGfNkY80piP5YnTvn0Ht5zUDgy/4n9JWO8WGmMucMY06UvEZGOwTDmdqnA1tpy/QAfADOCz/7QHw9ojCnYSEc7GvhLfzzLpmBzZaZZ3UAiH54hAzMS+W0NrARuGODn6VcYYyqAR3Dt3grYBrgCaO7hffPtPW8KZlhrhwOTgGuAi4Hbcp1ojNnkTbYHO6KOZMMY823geuAqYCywLfAr4PiBfK5+xn9if9F4sRewL3DpAD/PRpGHss/rMbdXZ1rGmKHGmJnGmDXJjO4lY8zo4JTJySyv1hjzmDFmq+S6qcYYG9znWWPMj40xLwD1wN3AgcCvE6//+uCeclT/mfz/VnLOZ5N7nWmMmZ8805+MMVsnn4uBPc8Ys9AYU22MuaavZ5/GmCuNMfcaY+42xtQCpxhjyowxv0gYgA+NMdcZY0qS8083xjwVXJ/GHBtjjjXGvJPIdKkx5sLg3OOMMW8k7+JZY8y04NhSY8x3jDFvAg192ebuwlrbBDwA7AJgjDnGGPOaMWaDMWaJMeby8HxjzFeMMYuTd/2DZKZ4+AA8ek+xI4C19m5rbbu1ttFa+7i1drZOMMb8NJn9LjTGHBV87tmfhBl7zhjzc2PMWuBe4NfAgUkfqenndnUb1tr11tpZwBeBrxpjphlj7jTG3GSM+Ysxph441BhTmsjmg4Rh+bUxZgiAMWa0MeaRpD+sNcY8o/5ujLk46Xu1xph3jTGHDWBzNwVRRwIYY0YAPwLOsdY+aK2tt9a2WmsfttZ+J9GL640xy5Kf640xpcm1IxO9WJ3I6xFjzITk2E+AjwM3JvK4ceBauen4T+wv1toPgb8C0zJtvzHmcmPM7zd2D+OIsUuTcWSVMea3iW5hnM9ybsb5bxhjPpP8vZMx5olEVu8aY74QnJcl+15qdq8iX8fc3nbK/gsYCkwARgFnA03B8S8DX8XNdocB3+7iXqcCXwMqgJOBF4AzE2byAoDEmFQmxvng5LqPJuf80RjzKZzx+hyOcVgGZLKax+NmYvsk532lG+3eXJwIzARG4AaGy5Lv3w3YEzgI+N4m3usO4OvJLHo34GkA40KAvwFOx72L24E/m8QBTnAScFTyHHkHY8xQnKF9MfmoHvd+KoFjgLOMMSck5+6CY09Oxs0KR+De+WDEe0C7MeYuY8xRxpiRGcf3B94FRgPXArcZY0wn99ofWACMAU4BzgReSPpIZd88ft/BWvsSsBTnPICzKT8BhgPPAv8P58TtAUzF6cBlybkXJddW4WzQ9wFrjPkIcC6wb9KPPg0s6ofm9ARRR9JxIFAGPNTJ8f8GDsDpxe7AfqSYtwKcHZ2EY2EbgRsBrLX/DTwDnJvI41wGEf6T+otx4eqjgdd6cJvTkp9DgSlAOYku4MbsLwXftwtOZx41xgwDnkjOGZOc9ytjzEeDe2fKPu+Qr2NubzuqrTjDODWZ5b9irQ0XSd1mrZ1nrW0A7sd1js5wu7X2nWRW3NbJOcfgZlCd4WTgVmvt68lM4RLgEM2WE1xjrV1nrV0E/IJAEfsQzyYz/Q5rbWPynJdba1dba1fhnOtTN/FercAuxpjh1tq11tpXk8+/AfzKWvty8i5uTz4Pc9j+11q7NHmGfMKfEiZnA3AE8D8A1tqnrLVvJnKbjWPaD0mu+RzwsLX2WWttC87Y2hz3zntYazcA03HP/xtgtTFmljFmbHLKYmvtb6y17cBdOCMxNvfdWGatvcFa25aH77m7WIYLdwP82Vr7nLW2Axf2PgO4MOkLtbgw8EnJua04WU1K7Moz1loLtAOluH5UbK1dZK19v19btJmIOpKFUUB1F2PFycCPrLWrrLWrcWkSpwJYa9dYa/9orW1IdOYnpOzKloAtvb9ovHgWR9Rc1YN7nQxcZ926mTocYXSScSkxDwF7GGMmBec+aK1tBo4FFllr70j60avAH3HjkuBln/gj+YS8HnO77agaYwpN+sKh8cCdwN+A+5KwwDUmPedpRfB3A2620hmWbMJjbCw/dTywWP8kxn0d6V5/+D2Lk2v6Gplt25rgOZO/N3VmciJwHPCBcSG9/ZPPJwEXJ2GbmkQJt6bztucTTkiYnFLczP1pY8w4Y8z+xph/JCG69TjmR6kl4wnak0yG1vT3g/cWkknaadbaCcA0XPuU8rIiOE9pG531pXx9xz3BNsDa5O+wfVW4iM6/A51/LPkcnPGdDzxujFlgjLkEwFo7H7gAuBxYZYy5J7FneY2oI2lYA4w2nefYpo0FBLbeuJS1m5MQ5gZcGlmlyb88wu5iS+8vJ1hrK621k6y1Z/dwspVLT4qAsYkj/ygpR/4kUhHaScD+GePtycC44F753M/yeszttqOasHTlwc8ya22LtfZya+3OuNn+ibiX1a2v6Op/4/KLDsI5xrnOBzeTnBRcMxwYCXwYnBOubts2uaavkfmsywmeM3kOPWM9zpgIoeJjrf2XtfY4XLjhEeCe5NAS4IqkA+tnqLX2vi6eI6+Q6NiDuBn8dFxYZRYw0Vo7ApdLp3DmclzKCQDG5VmN6t8n7htYa+fiJoHTNnJqzss38v+gQpLSsg2p0FnYnmpc2Pajgc6PsElFEWttrbX2ImvtFGAG8G2T5NZZa2daa6fj+qHFhUQHDaKO8AIuzeyETo6njQWk2/qLgI8A+1trK0ilkcm2lVuCiQAAIABJREFUDEZ5AP/R/aXLcbML5NKTNtwCI3CM4peMMQcCQ4B/JJ8vAZ7OGG/LrbVnBffKez3K1zG3txdTfdK4pO0CHIXcimtwb2AlLmdEOAR41VpbD07AOG8+POdu4OvGmN0Sx/Zq4Blr7dLgnO8aYyqNMdsC5+NyRvsbdwOXGZe8XgX8AFDi9xvAbsaYXRNF+KEuMsYMMa4cS4W1thWoJSXvW4BzjDH7GodyY8yMJJdmUCB57uNxk4t3cLk9a621TcaY/XA5P8IDwAxjzMeSPNwrSHWoQQXjkvIvMqkFHRNxKSkvdn3lJmElMCEjVznvYYypMMYci5uI/d5a+2bmOUk48zfAz40xY5LrtjHGfDr5+1jjFm4anH1qx+V5fiSxXaU4Z6eR3rNbfYKoI+mw1q7HhR5/aYw5IWFJi43L370WZ2MvNcZUGbfA9zJSNnY47p3XGLfA94cZt88ce/Iesb/wOi5kX2yM0fqTTcHdwIXGmMnGlbm6Crg3SCn5C86R/VHyeUfy+SPAjsaYU5PvLE7G3p17r0l9j3wdc3s7R3U88CBOqd/CsZ1399K9r8fNZGqMMdeRO+z/Q2Bmcs5nrLWP4RTqIZz3vy3ZDO/DOKV+LTnvzl563s3BFTiH9E1gNvAvnFONtfZtXGd5Crc44p8Z134VUMjq66Tyrv4FnAXchEt3eA+3UGIw4GHjNoDYgMsX+6q19i3c4rwfGVct4TLAs8PJ8fNwhnk5zmlfRQ/L9QwQanELXP5l3ArRF4E5OOanp3gS1zdXGGOqe+F+fY2Hk/e9BLcg5jrcos3OcDEuXPli0if+hmPLAHZI/q/DMXC/stY+hQt3XYNjmFbgohPf7/WW9C6ijmTAWnsdboHupcBqnM6cC/wJuBJ4BWdf3wReTT4DN7YMwb3/F3Hh7xD/C3zOuIoAv+jjZvQUsb84/ADYHjf2XYFjBjcFtwO/w42zC3GO+Hk6mOSjPggcHt4zSQv4FC4dYBlOLv8PJ6vBgLwec43LjR58MMa8BxxrrX2vm9cX4RjfydYtpIrYgpDMhmuAHay1Cwf6eSIiIiIiIrZU9OWYOyh3rDDGlOEqCHTLSY3YMpGkNgxN0ht+imNOFg3sU0VERERERGx56K8xd1A6qtbaJmttviVvRww8jseFXZbhwlYn2cEaMoiIiIiIiMhv9MuYO2hD/xEREREREREREVs2BiWjGhERERERERERseUjOqoRERERERERERF5ic528dhU9Dhv4C9/cRWmjj766C7PW79+PQB/+5ur7//Zz342+2GSNAbT6ZbWWejtml89lsezz7q6zHPmzAGgtLSUwkK3QcqOO+4IQENDA+vWrQNg+vTpAP7/cePGUVnZ7e25+10e1tqs99XS0sLixW5zkI4OV6Zu7Vq3scqGDRtobW1NO7+jo4OiIqfKutewYa5c7OTJkykuLgacbDLR1ubK4+n6DOSdfgww8k4eP//5zwGora0F4LrrruOAAw4A4DOf+QwA77//PiUlriSo+sno0W5zlbPPPpsxY8Z09+vzRh6d2b61a9fy97//HYAJE1xt7oaGBm8j9t5776z7bIb9zEReyKO9vd3bzEysWbOGP/zBbSa0886uxOXcuXP58EO3v8o111zTna/sDHkhj4aGBhYsWADg29ne7sqeFhYWMnSoq4v/r3/9C4BjjjmGf/zD1bHfaaedACgocJzWAQccQFlZWXefPy/kkQt33+2qaL7xxhuUl7sN3PR7zZo13v/4yU9+AsDw4cN742v7ouboFjnG9DRHdbMufv99tx3wz372M/79738DsHChq2KggaOwsJDdd98dSDkp77zzDtXVrpyfnneHHXYAnLG5+uqrARgxYoS/Th1rI8i7jvONb3wDwA8uO++8s5fbtGlu05nhw4d7x+orX/kK4Jw7gLKyMj72sY919+v7TR65BtbHHnPlCz/44AM++OADAO+w1tXVAe7dahCSA9ra2urvo8/0/ocPH85ee+0FpHRmypQpbLfddjmfJ+OZBlQ/6uvrAXj00Uf9QPPcc88BsOeeewJOPxYtWgTgHfh9992XZcvcpjuSaVWV2xVxr732YuxYt+X7McccA7CpfQXyqL+88sorAHz84x8H4MtfdnWoS0tLuemmmwB45pln/DmyKUcccQQAt956KwBnnXUWV13V7a3BB0Qesoub8t7OPvtsZs+eDcBWW7nt3keNGkVTk9tqXAP0xr5vMNjTruQiJ+yUU07xNuITn/gEAMuXL/f96jvf+U7a77SHGWREyI9//GMAVq1axZo1bmdLTVCWL18OOPvx+uuvA/jff/jDH7jhhhvSzpfjes455/D4448D8IMf/ABI9cFNQF7Yj6VLl/o+IYf9yitdSd3W1lZ23XVXAH77298Crs0abxsb3e6s0p2pU6eyyy67AClyZDMQHdVsDJyj+sILLwDwta99DYBFixb5WVlFRQWQYrS22morRo1yu3DJmFZWVnpHTAO2jO6IESM49NBDAadQ4BRmE415XnScEN/85jcBvHyGDRvmjahmt/vtt583KnvssQeAd04LCgr4yEc+QjfR5/LIZew1WMohX7JkiTckQ4YMAVKGtbKy0jsdL7/8MpAyNpBiXrfeemt/ve4r9vnoo4/2f0+ePLnT52KA9ENt/elPfwrAyJEjmTTJ7epXU1MDpJjglpYWXnvtNcCxzZA+cMh5lXMa3l9G98ILL8zJNudA3vSXt99+G4DDDjsMSNmRk08+2evDqlWrAMe2SiZ33HFH2vW33XYbn//857v7GHkjj7lz5wLw17/+FUg5Zq2trT5qJRva0dHhHZAjjzwSwMvgsMMO8xP+biBv5PHrX/8agPvuc/XJ5Zx2dHTw0ksvASnHwlrrJ3JyOt555x0ATjzxRL7/fVfLXqz8ZmBA5CH9P/300wE3VirioPf95JNPArDtttt6Oypn9tprr+WBBx4AUmOO9Orwww/noYce8vcF+P3vtcHXRjEg8njzTbcpl6Kxzc3NXv81Vr711luAI4hEbIwcORJwEzsRJrIzYluXLVvmfQ2NVWeeeab/eyPoV0c19IlCRj0TYo/33XdfwDHyIhIlt4kTJ/KLX7i9LySnXkJOmcQc1YiIiIiIiIiIiLxET3NUs5DJTNXV1fm8GDE/q1ev9n+LCfjSl74EuNmLrlVI/4gjjvAzH7Gs48ePdw0oKvKz5v/6L7dT3H333bc54cy8gHJTNbNXnt3rr7/u2bCwTZrR6bOGhgYgdx5mPiFTP5YsWeLTOsSq77nnnn4G+4UvfAHAn1NWVsb5558P4FnGwsJCz8I3N7vd28SgFBcX+9ngG2+8ATgZizUSo6rn6WGOXq/g0UcfBVKpCsOGDfPt1/OKPW1tbeWkk04CUrP9BQsWsGLFCgCff7btttsCsHLlSq9bkvGsWbN8yslggdiBzIjQddddx2677QakcjJbW1t9eF+RCjEJYpoGE8SEKzz99ttve30Wgx5GGfbbbz8A5s2bB7j0CLElsqe6V1VVlbetSjX61re+5ftXPmP+/PkAXHzxxb5/iAUN2VDJSvnKdXV1vq8J22yzDeBSbY4//nggJaNPfvKTfdWEXoF0Wn1jw4YNfgxW2xWZGT16tGdSNfbMmTPH22Lph9jnlStX+miO+mA+o7a21kcUZDsLCws946n0qn322QdwOduKRmhsXbNmjc9jl4w0RoTMqSJVt9xyC9/61rf6rlGbiVxR81z+kaJMH/3oRwH49Kc/DbhIpOSkiOXvfvc7H+lVlFvYjFShTcbg8uYiIiIiIiIiIiL+Y9DnjOrChQt5/vnnAfjnP/8JuHyo4447Dkgt6BDD0dTU5FmPU045BXCsW+ZMRh7+bbfd5vMRNSOorq72LFo3EuAHBJKRZixa9d/a2upZsbDt+kwzXi2oKSws9GxAPkHvIXOmtXLlSj8j04y3oqLCM6PXXXcd4HJiwM1gxaiqzdZaf1+x6+eeey4A22+/vb+XGNi6ujrPOOZ6zoHWFTGqyr0eM2aM13cxIWKIiouLPVOsPlRVVeUZVOWR6brKykqvM2rn7NmzN1b9IG+RyeqMGTOG995zOytrwVVxcbHPo5KM1HYx1YMJij7JTk6bNs33L61Gln7/9a9/9f1qypQpgGPHxDKpX33uc58DHFsrVlbRrjPOOIMHH3ywbxvVC1De5Zo1a3wUSuyibMD48eM9OxiyjFOnTgVS/UT9oLKy0t9DDFK+M6piitX3rbWe+SstLQVSrHxlZaUfW9XOlpYWf57y+6Vrzc3Nnq2XHdmwYYOP5uQbFi1a5Flk/W5vb/fPLnsg/aitrfWMomxLUVGRZ+hlh8PcTtlOsa7V1dX+fpLjQEK2Lowa5vKLlO+vSJQW4ubCzTffzPbbbw/ApZdeCqQWpPVFNDsyqhEREREREREREXmJXqdPMtmoESNGcNBBBwGpPMrdd9/de/QrV64EUqvJFi1a5Ge8youqqKjw91U+jY7NmDGDJ554AkitfF+7dq1nVAcLtOowMxesqanJsySaza1fvz5rFix55iObCqmcuUzGbtmyZV4HVNfQGOP/Pvjgg4HUCsQrr7ySyy+/HHC5aAAzZ870DMGNN94IpHKq6uvr/TFh3LhxPsdXFSnEqFRVVQ0oC7969WpfOkvMSEFBga+hK8ZHzzhs2LCslcqhDoTMEDi2ROyAMHLkSJ9fJYYt3yG2Q3olFjBkj8Q652I/Mu3PYMGCBQt8u/SuSktLvW2VPMSi/uQnP/G5mzrW0tLCgQcemHbfkAFStEY2dPHixX7ltEr35CPEIpeWlvoKB7IDilQ1NDT4HGbp/NixYz0jqP4VlhrS35l2JF9x++23AynGs7W11Zf3U9k/9Z/333/fjyH6vWzZMn+eykd+6lOf8sc0HkmmM2fO5Mwzz+zbRnUT9fX1WfmoxpisCg4h2yhZ6bPS0lJvX/SZWMP29nbfr/RZU1OTZ+sVxcgHhCv9M5/5hhtu8P3n8MMPT7suzDkNI2+qOHT99dcDKUa1L9Dnjurbb7/tF04tXboUcI6WBleFl5TAX1JS4il0hZtWrFjBCSecAMD9998PuHQAcCWKFNYTtX/zzTfzs5/9LOfz5Cs0mGpwkCPV2trqDYMWQ6xdu9aXaFL4W46rDHQ+IUznEMK6fnrfodMtB0sdROVS1q5d6x1VYfbs2T7Ur/ZfdtllgEsV0CCkckUrV670G0wojHfXXXcBbvGWOqMWZPUn1Ecg5TwsWbLEP4t0Xb9bWlrSSoYIMsQyuqFhVaqF7jFx4kTvsA0WR1XOlPREjmd7e3vWoBKmhshA63xNBgYLlixZkpX+VFBQ4OWhdmaW8wsxbtw4r1uZzldRUZG/NrQlg8FRXb16NeD6beZERnZ17dq1fhIs537UqFFebpKj+lRDQ4O/h8acfIfGEpVZmjZtGrNmzQLg4YcfBlJlqu644w6ffvfII48Azk7KZh5yyCFAKq3i2GOP9WSKSiTm80K7DRs2ZKWdFRUV+cmHdDx0TmX/NQaHi6/UX0L90rglm1tRUeEXquaToxraw0zCSH4VwKmnnpp2rLW11bct9KfOOussIOWnafOVCy+8sEuyR7LLTEnoCjH0HxERERERERERkZfos5UTmpVcfPHFPqyqUMvVV1/tGUCFKVV+qrq62rNtKpkThmEUslI5p7vuusuHzY899lggVYZoMEGzMiUyi2Gurq72bJdKaPzqV7/yMtGiAS0oy0eUlZX5mejMmTOBVDHqW2+91S/8CbdGFYOoosJq+4MPPuhn/lpUdeqpp/rtMsX8/Pd//zfgZtFiGDR7fvXVV5kxYwaQWnQVMokDwaQKc+fO9TN/hY8qKio8M6b3LTZt2LBhPgSnxP2GhgZ/XLoTllPJDHdPnjzZMwCSc74jszxbWMxa7zss0bSlYN68eV4/w9JBmaxEuNhOqR5iEMMNUTK3HrbW+n4Ssq0KieczVKi8sLAwi7WRnowcOdKzySpmv/3223vbk7koKFyspz6V78iMOEGqdJdC+rJ3Q4cO9ZFIMaPjx4/3kR2VNhMje8wxx/jxfDCgsbExy0ZYa7OYUUX8jDG+n4gVDBfYyqaEZakUMtc4VlJSkpeRmjDaIvsgfXjmmWf8mKzdLoVwQViYRqUUAclOCzQvvPBCL68wVaAnKXWRUY2IiIiIiIiIiMhL9Bmjqpn6rbfemrXveDjLydzus7Ky0ucCiVHacccd/Uzm3XffBVIFzxcvXuwLU59xxhmAmx0MpnI7ra2tfkavnBbNbhcvXuxZZ+UL3XLLLVksq2Yrmbmg+QLlGKudWgB34IEH+i0KtYho2LBhfsYqdkALhR555BEuueQSIJXMP2bMGE4++WQge3GMMSardM+SJUs8g/g///M/APzmN78BXEmj8847r1fa3B2otBKk2OEDDjjAP7uYLc1sOzo6/N/qZ8XFxZ5x1+xVM+hx48Z5Jk45z7vsskvOXMZ8hnLKxI5J79vb23PO2DOLXm9OflQ+YcWKFZ5NFmvT1tbmbURmKbvCwkL/7hWtKSoq8nKTndT5LS0tvo+KQSwvL/fyzmeo1NiQIUN89E79IMxBVfRK6ySKioo8i5aZ+1xTU+N1S8cGI2RvJQ/JwBjjx1bpyfr1672NEHuvnN2XX37ZM6qDofRjY2Ojb7N0vaGhwZfuElsYbpSh96zIQuhDyA5L10pLS31f0vfU1dV5+eUDMreTDxeShdFnjYGbC5V9DMvDyX8JN47I9f2bij7z4sLFCnI4RZHPmjXL10hVkruMYllZmV9NpsGztrbWK48MimjnW265he9+97tAKrl71qxZfvedwbD6v6amJmtVnYxHdXW1T3bXy1+/fr1PDQgHGEiFRPMJjY2N3onS811wwQWAqxeriYkG0lWrVnknVNcJP/rRj7xxufDCC/3nSgXRe1doq6CgwMs2rAWYGcq85ZZbANepB9JRXbdunW+fnIMw5CinVL/DfZvV54YMGeIHIi2Y0r1KSkq88ZCROumkk7yxHSzQwrjM8L611n+Wa+ecTId1sKUHrF271ht6Dba1tbVZjqQcjHAhmc4Jw5a6lwbW+vp6/3e4QCufBt7OoLFhxIgRPrQrvZZtqa+v93ZANmbYsGFeDrKjktH69eu9zshZy3fkciD1mRau6lhNTU3WXu257I3kowkApPpXrv3i8wUtLS1e/6UT7e3t/j2rDfq/vb3d60cukkv3UF/6/+ydeZxcVZn3v6d673S6kxBCVkgIAcISWQKIBGQzLLKqLCIDURkXxvF1GcVRFHCfeRWXmVdgFBREQQeVRQQFWUMiIDsIaPaE7EsnvSW91H3/OPd36tSt6k4n6eVW5/w+n/50Vd1bt+459znPOef3bNXV1W7DKJmprq4uyKwymCiW1/SPf/wjkHORq66udsF3WnfJlWbChAkFpvzNmze7OVZjTG4zZ511Ft/4xjeAXLByMQJtRzY6wfQfEBAQEBAQEBCQSvS7XXzixInONO+nv/jd734H5FbcP/nJTwC7w1OaA5mER40a5fJpKo2QzMStra2OXRQ7su+++7qArFJgVDds2OBYNO0y9L6rq4uxY8fmnb98+XJ3XLth7Qy140kToihyJnntPmVCefnll92OXIz7lClT3GuxyUceeSQAd9xxB3fffTeAy91XU1PjqpIlqy5BoamhmHn4oosuAnImssGCX3VMrGhZWZmTY33mB1OpfdrZvvDCCwUpeLSrrqqqKmj7ihUrSi6fqNraEyPqB0MUq3cNdFuhLK3YsGGDGzticLZt2+aeqeTfZyv0WrJQUVHhGCSNPTFAXV1drm/FIEZR5NiVNMMf+9KDyby5w4YNczpIOrSqqsqxZ9Kj6oOtW7e6fk6zibsndHR0uIAp9YPYr87OzgK3sebmZmf5EussfSKXKUg3kyr4eYHFhmYymTzXGMg9Wz8wUTLg5x1VX/luIDpPn7W0tKSCUZUb2Q033ADAnXfeCRTXeTU1NS6NqCB5kN6A3JiZNm2aGytag2icvPXWW67K1axZswAbVC8rz2mnnQbkB3Bub2wFRjUgICAgICAgICCV6DdG9ZJLLnGvlZxeVS722GMPV0VKjrjXXXcdYCsFaSWvigetra2urmwyOfwHP/hBvvOd7wC5FfoLL7zAAw88AMDjjz/eH83rUyxYsKCAFVMhhGIJ2E866STnI6NdjXZwaUy+XFtb64I9tPuSH8zDDz/sdvfyf/J36mLLxZi+853vdEF5zz77LGBlRqmqlKJM1dDq6+sL/GOqqqrcbyrx/+c+9zkA7rnnnj5o8Y5Dz2/48OFu1yrWtL293e2OtRtWdZ2Ojg4nM5KFN9980zHQcvqXT2d9fb3b+ft+5GJcdR9prd0taPwXY1STPlm+r6raLAbN97krBTQ3N7sxLgaxra3NjSs/qTnYsSRGRGOwtbXV6RufLQE7VsWu++mv/NRxaYXvl6v+kNzr/9ixY53MyMcXCplXXWvmzJm8/PLLQH4qtP6oZ95f2LRpk9OpySqHTU1NBamaOjo6XH9onKm9peLLrvvs6OhwMu77oUoG9F/PO5vNFlRR9BPk6xoaP52dnXnMoL5XzD9+IHHDDTe4oGPpdM2DI0eOdJXrFKQMuXRtfqousPOl2FJZWerq6grY5eeffx6wVp+TTjoJyK35Lr/8cmcVVHzSl7/85bzf6QmlM9oCAgICAgICAgJ2K/Q5oyoGR7XZP/7xjzvfCNVhP/roo10mAPnAyAcqm826sqDaFas4AMDhhx8O5KLHf/7znzuWVb5HF110kSsRVwrYtGmT27GpXWLQipVgmzlzpmM7dL4S96Y1PZXuS0mRtftat26dY7W0a1u1apXb0at06nPPPQfA1Vdf7WRGfsuQq28tPxtlCygvL3dyJHZ25cqVLgo46fOpNFoDDcnus88+66wH2rFms1mX6F87eb1vb28v8BVqbGx0n6md+l4URc6/+9VXX3W/re/qPtLOqHaXfL1Ycu7q6uoCRlAMUan5qEZRVMC4NzQ0uPEi5sz3r1Nb/fRNyTKzvl+eztfYqK2tLYlk99KhnZ2dLoL56aefBvJ9d8WeScb9wgbqF/Xx+PHjC3z3fP1RCujs7HTjWmy8778tFEtwr3m6Nz7haYKYT58BljzX1dW5uSZ5nnx5IT8Dj2TLZ2ohf3ypb7Zt2zZojKrG6ZVXXunuWSyoxoCfFcOfV5IZQITy8nInP2pXY2Oj6ytZdzWepk+f7q6x//77u/MlX8oI4BflSabQSqLPF6p6uMopVlNTwzXXXAPA+eefD8App5zi6GItMm+//XbApheSiUqL1/LycicE+p4U0IQJE3jqqaeAnDn5+uuvdzS20jDIgTeNWL16tXOB0MPWoJITuw+/lrCEUQ9Ypp20Qe4fEngpwjVr1riFqgZNVVWVG1yXX345kHt+1113Heeccw6Qcy955plnXL1hVVCR6X/06NHOdKf8f+3t7W4gqf/kaqEqZwMNLSLGjh3rFk8yleyxxx6uP5KBYcYYtziRebOtrc0p4mTd6tGjR+cFiuh89YPuQ/KYVvh16CHfvC/ZKjZZJIMo0hD00Bv4wRzSDXq2s2bNcjkMpU91bOvWrW5S9WueSwakN9Qfq1atcoEQ8+bNc9fyF3Npg9qldmazWbcp1Xjx8+xqoSq9U19f78aL+kHp64477ri8ij5gyZU0L1STptTW1la3QNVz1P/Ozs6CQLJMJuP6S3pHJtxSySXrm/STqcf89YTg91kyRaSfq7pYJbekq4BfGW2gIZfHiooKl6NerjyS6c7OTvfad1eQTtD40P/q6mo3f6gfOjo6XLvVr9rc1dTUsG7dOiA3/hoaGvI2vpCbq88777wQTBUQEBAQEBAQEFCa6HNGVeZEOen6qWHEsnZ0dLjgGJmCxWysXLmyoNb6ggULHBumlbdM4k888YQ7T2mtJk2aVDQIKa3YtGmTq5KSrDMshsTHHnvs4czf2tWIfUtryhDVEb7tttuA3P0uXbrUPXv9P/bYY933xIbKtaGurs7t0m6++WYg31Qls/a5554L2MAi1asW41JeXu7MFmIVdc3nn3/eFSAYSBmS+WT8+PH8+c9/BnLs+vjx4wvq18vEk81mC3ajZWVlbnebrGTV3t5eUDRg3bp1zoqRZubMR09MaDK4oVgwldqeDCZKK8R2GGOcHEufHnrooS71X9J856enkntHe3u707E6z2eY5V7117/+FSgskpA2aOz4LJksAuoP/c9kMgX95we/iBXzWbJk+0tljAitra1OzhVAJstWJpNxsuAXTJB1RrIgNiytFrskkvMo5Ji/KVOmOJnReX6lP50n5s9nD/VfrGs2m81LlQlW1sRID3SFTOnxjo4Opk2bBuBSdQqtra0cf/zxQG5eHT58OCtWrMi7V8l5U1OTszho7FRUVLi2aa7R+RUVFW5d57tYSP9obagUpYFRDQgICAgICAgIKFn0+TJfpU31H3AMlfDRj37UvVbZ0wULFgCWARDb9swzzwB2Ra/dgT5THfSXX37Z7Q4/9KEPAaXjRyPU1tY6fxL5Avk1uJOlxkaNGuWc3JPO/2IJ0oaZM2cC8OijjwK5nVZNTU0Bq9Xe3l5Qy13vx40b587zWUXt9H/xi18AOVkYNWqUY6vFLra3t7s+1fX1/dbWVpfSTGk0BgLqg/r6escW6tkedNBBBWPI9ydLppDxk1Yn0+5s2bLF+dfpnLa2tgLf6LQjyRL47KnYkZ6YwN6ckybId7i8vNzJhfwO999//6IMkiC50PeKpZryy6wqEMlnOdKcjkl6oJhuSPrdZbNZxyKLLdy4caM7P5ns/ZVXXnGfiWkqhcAyHytXrszz34V8/3TB9/XVeWLGxCQ2NDQUzEe9Sdg+0PBlXM9efpbjxo1z6wgx7X6wlF7re+3t7W58JS1VXV1djp2U7qyqqnL9p3lroMrv+tZIBRRpyVYFAAAgAElEQVQn1wT777+/m38VywO4lFWaMyQb/nyifh0xYkRBcR3NIWPHjnXt1tpsxYoVTofIUinr6o033phX1rkY+p2P7urqynPABWuqUmDVr3/9ayBXmcqvnauAmIqKCqeok1HZyi4A+QvUHakjO9jwnflFxauiAxS2ZerUqU5IkjXr0wjfYV/5Tq+66irARqWr7b6pQQpE+XZl0rjzzjvd5kamivXr13PZZZcBOcWjzc7YsWPzAkvAKl+ZIfwFIlh5VZTvQC5UfSWnRan6pba21ik+KUrfxUNyr01OU1OTUwrqD990pddyeJ80aZJzJUnW/U4rkouFYgs0/zP1n/77ekGLEvVtGqHA0kwm4+RYC9X6+vqCinR6xv6iQzq0urrayUdyAvI/k3tMY2Oju34a+8oPegE7HjSGk0Et7e3tbsEyf/58wOoW6SdNsJrc161bV+BCUiq5RIXVq1cX5E/1Mx74FZsgPxgoqSez2azTG8Xc0tKGjo4OJ/fSkzU1Ne4ZykTvu4aIKPCfezKIVe83b97s5iYF5PqBnlrLDNRC1YfaKGJCZGBVVRULFy4E4B//+AcAxx9/vHNhUKC67nnPPfd0c4ue+erVq91YUR9qDL300ksFriIvvfSSuy/Nzbr+jTfeyKc//eke25LebXJAQEBAQEBAQMBujT5nVJPsn28y8gNexBDI/CJTnr/j9YOq9Fo7XTFK733vewt+06/lXgqManl5uWO0xJaIVXv7299ekGOsoqLC5drULk4MxzHHHJM61wc/IEH3KUb1hhtucPerXfvf//53t+O//vrrgdyu8NFHH+Xhhx8G8nNGfvGLXwRwOXi/8pWvAHY3rJ2eH8ynna7+qz8hv1rHQMHPeZdMRdXV1eX6SPLhV4pRu/z0IIL6VN/v7OwsYN/86lZJ5iCt6M6S4OsbnxFJMq6+XhDjImYkjfBT64gVk0uLfzxZkzyTybjnrXzU27Ztc+2XfCQD0CBnKl27dq27vtgmP7f1YENjws8lm2TRNKY6Ojrc+X7AlNovBkiM0JgxY1yqK6X5KYUqXT4UGAU5fSc3Cb+ykuCn75MM+BYnXS/NjKrv8uQHiYGdB5KfSXdqHoacLmltbXX9kMwx6o8l6fAtW7a4uSl5fn9DTCnkLGa+SwxYy9vs2bMBXAXD5uZmx5qqTxSQ2Nra6uYHPXvfCqE2ai1SWVnpmHiNwyVLljg3Eo1J9flvf/vbwKgGBAQEBAQEBASUJvqcUU3uzvz32q1XVla61bjSVInNyGQybjei1fj8+fOdX4OchcXI7rvvvm6H6/uZlAKTKlRXV7tEvdpl+DXIkymn1q5d6/rm4IMPBnJpnPzKGmnGN7/5TcAyqkkWY+PGjY4xk8O1dsBr1qzh85//PJBr6+rVq/OcyCHnSD5lypSCwIGVK1cWBGvJ32/YsGHu2EBC91FeXs6hhx4K5NgrP7WJxpAYLj84RExbeXm5O67zfRZV/S0WedWqVW63O9g1qnsLn/noDtvzWxVKIThGsmuMKaj64iOZyDybzbpnqmv4fsqSC7FHfv+MHTsWsDpaulXjMk2MarJqjj+/SKfIIlNVVeXGt9ikxsbGAgZR42f06NGOWVL8QJoDy4rhxRdfdLpSz90PYPXT9kF+wZBkYYjq6mqnWw855BAgnVZLtcn33/YDjiUres56/k1NTU7ufcZd5yWLiWzatMmNDbH2LS0tBWzjQEFFjyBnpZZMq13Lli1zBXKk+zo6Otwz1lrroYceAuxY1/wo1nTcuHFORrQWUX8NHz7cMa/SUTNmzHAyl/TbVUrTnlBaIy4gICAgICAgIGC3wcBkoY3hp3vQSlvsjvwostmsW3Ermmz16tUujUIyavHII4907Ip2DH6KmlKAX6tbDIB2NVC4Y62urnZRrTNmzABydezTuLuFQt9l/d+yZYtLzaW2t7W1ued89913A7gk+CtXrnRsjko8nnXWWS45uVKFKIKxoqLC/bbkqqWlJa+0G+RHxQ9UcuZiKJaQ3y8DmEx70tXV5caQ7ruystL5ASUZttbWVscOqGTtwoULnfzIUpF2+AwjFGdKk6m5fPjjJFmONY1Qe4cPH+58z3yfWh0Xk+RHwqsfxNBXVFTkyZZ/vi8v0i233357QYq3NEFjQc95r732cm2WnvT9VyXjaovPkCbjAZqamtx1xSr5Pp+lgPXr1zt/0uRzLi8vd6yfGMfm5mbn965j6oPq6mrnq5tmiAmOosjpfaVF8ll1zRPSnXV1dW7ukV7o6OhwrKTkQuuPxsZGN+bUx7JuAgXxAP0N+Zn6vy2/Vc0J1dXVbvyLFa2srHRtVN+oOERlZWVBae2mpiYnG8nMCTU1NW5Okpz5cROa5zWeeiNPA5qeSrjiiiu48sorgVwnauK48MILXRoFVZ+aNm2ao4mV5kCdc9ddd3H66acDuYVqqWHEiBFucGjgqN57MURR5AaRzDAahGldqEpgNdDffPNNwN6vhN+vVS9Bv+aaawA455xzAHjsscdcG7/whS8AcMEFF7g8rd/61rcAuPrqqwFrntK1pLw2bNjgNjcajFLSHR0dgyJHGry+GVrPuLGx0fVbUvFVVla6yTWpkHz4GwQ9C43L2tragrrNaYcWXcXMsMnF6/YWqhprMmWmEZJdP3jDr1+fbKv6xXflUIBDMRlLVj6DXA7NTCbjrpvGqkzJVFtVVVVuE+b3G1gzpp8vFOxiXWNGC3ed09raytve9jYgF2SZzGmcdowaNaqgEluxjYkWa62tra4SpJ63+qe8vNxtlNIMtauqqsq1WZtx3zVQel/6t7293Y0rP/euv9EB8lIq6re08KqqqipwERgo+MGwaofSOWohWl5e7tqrc6qrq92zVhslD52dna6NxYgMjSPffSKZ9s4YkydD/rHepLoLpv+AgICAgICAgIBUot8Y1WIshlbco0aNcqt70fKXXHIJACeeeKLbwfopQ2SmUQJ4MQGtra3OdOz/dikl/N93332dk7Kc/rXTKYZMJuN2J37amjRDu1OZFd71rncBMH36dLe7kzmlurraOYKrXXIB+N///V+3axR7eumllzqmXRXLFKy1atUqxyTJlJnJZNwubvr06UDOsXvdunVFg1T6G9qNv/baa+7eFFS1adMm1zdiB/Tca2pq3O7eN91I/pNyUVZW5sahrv/iiy86J3xZJ9KOZFqlpMnWR7F0db6VpyfrRdoQRVEBY+4zfH6KPshnj/Tfd4fxa5zr+oJcbPz0XmlkVGVVknly9erVziqice2nN/TT2oFlC6WXdEx91d7e7q6h6/cmkC8NSFa4g9xz9i0zeq05p6WlpYB1lk6qrKx01rA0wy94IV3puwaqPZILPdO6urqCoiC+/vCvC1bXimH2Xdg0RyWrLvY3/N9LFqrQvW/bts09c79wQZI1lvxAbh7xdUl3a6vy8nKnh/Wbf//7313/y8IpdrY3BTQCoxoQEBAQEBAQEJBK9BsNVyzhv3Zsxx57LJ/5zGcAuPjii4Gc/wjkpxkBW9JLOxjtErTbnzlzZsEuP+3sYhJjxozJKx8KuV1NU1OT24EI7e3tBbuSNJU0LIZbbrkFgC996UtAzim7tra2IIAsiiLHkiT9RS+88MKCBMZ33XVXQdCQ7+srxknyt9deezmHd+0adQ9bt251/q4DCfnUtba2Oh9CWRb8dEJqg5/2RDthP2VbMq2MUF5e7hhE+SBu3brVsQ7JutBpRbI2u1Asub9fwle6xWdUS8HnTs+4q6uroCjD+vXrCxhR3w9ZrzUO6uvrC3xS/f5IBklkMplU+y6rjLb82GfPns1rr70G5ORD48Fvhy8LyTLf0qfr1q1zpbznzJmT93tph/RBWVlZwZjwU/AlWdZt27YV+Kz7wUcKbE0z/GAqybPWDH6wWDLgsqGhoSCYO5PJuFgGXVfWqwMPPNDNxdK15eXljqkd6OIQPqPqW1C6g56rH4BbLMVob1Ky+VZsna++KSsrc4HL6hP1fW9Y535f0fmN1g1mMhkXpa0JWpPn2LFjXQ13mUOPP/54t5C49dZbAfjEJz4B5FdUkIk3iqKSMPkLw4YNc6bdt956C8hFoy9atMgtWIT29vY80y/0LIyDje9+97vce++9QM78roVGW1tbnpM3wIoVK9zGxc8LB3Dfffe5RajQ0dHhJiZBNYz9c5Wb9s0333Ry9N3vfhcgLzfrWWedtZMt3Xnoee6///788Y9/BHIZDDo7O93iREpUsr5hwwanbCQD++yzT8HglxLxzZxSTHvttZfbJJRKHXOZYXub3SNpzvNRCgtVLTqiKMqLoAW7sEwGbfgmOk0aikpuaWlxk3Cxil3aJCqYdcSIEQVmxDRB+YAVQAm56OfkgmHz5s0FOWT9am3SQeqzDRs2uLH2sY99rF/b0dfwA4SSuVLlSudnWPGD85LBr9I/dXV1eRlp0o729vaCAMO6ujqWLl0K5OcRBTuWilW6lL5JkmKLFy8uqjOTpu+Bgk/s+FkLkveUlPPOzs5uc776ue19ArIYSSAkA1o7OjryMtVAbs3Xm4qAwfQfEBAQEBAQEBCQSvQ7o+qzm1qVr1271lWkSqYJWrNmjVvZaxfz9NNPc+KJJwI4s+w999wD2OAZBdD86le/AkojgCoJP7UF5Niuf/zjHwWMakVFhcs95lfDSCtOPvlknnzySaBwl1dZWekCp/S8IccYyjx9wgknAPDAAw84Run8888HrClOeXYvu+wyIBeU5ptJ1WejR492ZkIxrz/4wQ8A60YwGBCr19XV5WRBfbBlyxbHbvkpU8DubMWeqq3Dhw93/ZxMt1NeXu7Gmsbjscce6xgGMbZph1xHZM7uLnjMPwY5Fs1nWEshmMo3lyUtCj5ro+fuV5ryKwLqWsl+8NNbJdn4MWPGOP3iB1ikBcnAn8rKygImy0/XlUzl5X8mdqyYi4hQLOViGuE/Kz17mbzFtu69994uLaDMtKNGjcpjY/3vr1q1qiTmVz3PqqoqZ530n5lSYKqdSYYRCseSDzGrLS0tBbLQ1dXl5vFiQeX9iWnTprnXugfdfzF3MR/duSv41e12BVrraa6W++fnPve57X43MKoBAQEBAQEBAQGpRJ8zqj2lhRKL2tbWxnnnnQfkWFCxPFOmTHH+mkuWLAFg7ty5nHnmmUDO+Vc+Nvvss8+A+4H0B5IJqgWfZRSMMQXpZXpKZzXYOPzww939aZevZ7xgwQLnOyd/0f/zf/5PQWJq1QMeN26c292JWa2trXXyo12hrr9161a3oxQzfd111/G9730PyDHzSYZ6oCEGfdSoUc56ICbAf7ba+atfJk2a5NhYMSLDhg1zcpSsaFVZWenGppjskSNHus+SgTppRXLX77MB6qNifttJ37vq6upUVltKQn7727ZtK3hGTU1NLsAwyZT4yfrFlvv9kiwM0NXVVcAgVVVVOR/ZgU630xsUSzmm8VQs/Zb0ho5VVla6diVT+RTzn+tNYEkaIF/jzZs3uzgA9Yv0XTabdayi2tXe3l5QiclPkC8GVikVxZClCWpnc3Ozs1D5UHCvLDLFfNjFSPrzra9j9T651mloaHDypLltoHDMMccAxZnc559/HoAjjjjCyYaY5eOOOy7VVoLSGHEBAQEBAQEBAQG7HfqcUU3uLnwfVbGG119/vWO85Pu0bNkywEaCyTdEDNiIESOcf4VYVkV6VldXu0j5Uoba8+CDDwI530z5DvpYsWKF27Gp7Sp5l1ZceumlQC49lXarkydP5tFHH807993vfrdrl563mFg/rYqYAMj1l1hC7Q5HjBjhEttPmTIFsH2sHfFjjz2W99uD5X929tlnu9diC7/61a8ClsV67rnngFy/iWWtra119+vLTLHE3mDZErEk2lXX1NRwww039H2j+hFi+MSeSj+Ul5c79rEY5N+p8dPR0eGYoTRDDFhTU1Oe3IMtgqKId7GFev719fUFqf38srti3tUfbW1tTo6ExsZGZ9l55JFHAPjgBz/Yh63rG/gJyyUfYst9RlVzjm/hk/yIMdO1khkWSglHHnkkYOdUPV89b7HrxhinY9X2bDbrxtDDDz/srgH5aYw0n6cRuv8nn3yyaOpGWav0vy/xyiuvuP6VX+bs2bP7/Hd2FEcccYR7raw6flrQNGNAE47KPPv88887BaIHqEmzubnZTSaanNesWeMWcDJbSXG+/PLLLnDGRylVpgI444wzgNyCTH1VzPR0wAEHOFeIww8/HMgppbRC9/ub3/wGgH/+538GcpXGfNTW1jqncN85vK/gV1/SpkgTWRrSfOkevva1rwF20lTAoDYufmW2pDm2srLSKWeZ8GQurq2tdSlMtDlS0FYp4fLLLwdyilbtPfnkk/nhD38I5AIvx4wZ49KRve997wPgxhtvBOxYOvbYYwfuxncS3/jGNwCrJ5UPUhg1ahTz5s0D4KabbgJywXnbtm1zi3ktcCsqKpxpWxs2jbMLLrjAyYxw6aWX8vjjjwMUBHamCf4G86STTgJy84RcesrLy93iQWSJn2tWCzktZlW5zkepzCkKOHz99dedPGjhqdywF110kUvlpep+s2fPdvPz/fffD+Sq2J155pmDkr5vR6EqUQceeGDR3NDdBTn5n/vPOZluyUdSHs444wy3+T3kkEN28M4DiiGY/gMCAgICAgICAlIJM9DpEwICAgICAgICAgJ6g8CoBgQEBAQEBAQEpBJhoRoQEBAQEBAQEJBKhIVqQEBAQEBAQEBAKhEWqgEBAQEBAQEBAalEWKgGBAQEBAQEBASkEmGhGhAQEBAQEBAQkEqEhWpAQEBAQEBAQEAqERaqAQEBAQEBAQEBqURJLlSNMUuMMacO9n0EBJQCwngJCNh9YYyZbIyJjDHl8fvHjDFXDPZ9DTSMMXONMXO6ObavMaZ5gG8poJfY5YWqMWaWMWaeMWazMWajMeYpY8xRfXFzQwnxYqHNGNNkjGmM++xjxpiS3Cz0B4wxlxhj/mqMaTbGrDLGPGCMmbWL10yVUg7jJR/xs9ZfNh4jev+Bwb6/NCHokO1jqOsQTwaajTFrjDE/NcbUDfZ99RcGSj9EUbQoiqIe+7G7ha4x5gRjzBPGmPJ4QzC5r+5rIJCQqU3GmPuNMZMG+7587JKCM8bUA78H/gsYBUwArgO27fqt9T+0wxxAnB1F0XBgH+DbwFXAzcVONMaUDeSNDTaMMZ8Bvg98E9gL2Bv4EXDuYN5XXyKMl0JEUVSnP2AZdozos18MxD3sKAb5HoIO6Qa7gw6JcXY8Xo4AjgKuHuT72S52VhZ3VD/0B4wxme1sBs8E/jAQ99KPkEyNA9Zg56j0IIqinf4DZgKN3RybA8wFvgNsAhYDZ3jHG7AKdhXwFvB1oCw+NhV4BNgArAd+AYzwvrsEODV+fWB87Yvj9+OB3wDr4s8/6X3vWuAu4HZgC3DFrrR/B/vK3bP32dFAFjgE+BlwA1bgW4BTgaq4/5ZhhedGoCb+7mjsoqcR2Ag8CWTiY1fFfdoEvAmcMlDt3Mm+aQCagQu6OV6FnYBWxn/fB6riYyPjflgXy9nvgYnxsW8AXcDW+Pr/PcjtDONlx8fI14FfAXfE8jwHqAZ+6PXF9UBlfP4VwGPe98uBCJgcvz8LeD2+1grg09655wAvxWNqLnCId2wF8DngFaB9kOSnWP8EHRLtVjokTwaA/xvfb/Lza4Hb49eT4zFQHr9/TGMZS1ZdDSwF1gK3AQ3xsQeBTyR+/yXgPfHrA4GHYtl5E7jQO69AFvtD/oucUwv8EqsLG4FngNHxsblYYmBeLNcPAqPiY/sBkXeducDXgPlAG1YH+XLwfe/cl4EZ8XWjuL3NwHvj4x8DFsT3dDcwLv5cuulfsbp3PXbzmRlkmToT+Hv8+t3AC1j9vxy4NvHdy2LZ2QB8uTfPaKfucRcbWB/f4K3AGcBI79gcoAP4Z6AM+DhWQZj4+N3ATcAwYEwsUB/1hOZdWOWyJ/BEQjCWYJXwEVgFfJY36J4DvgJUAvsCi4DTvMHbAZwXn1szWMLgfb4s7pufAZuB4+J7q8Yq03ux7Ntw4D7gW/H3voWddCriv+MBAxwQC9T4+LzJwNSBFPyd6JvTgU5iRVrk+FeBv8RysidWIXwtPrYH8F6sghoO/C9wt/fdxxjADUkYL307RrAL1XbgbN0DljGbF7d1DPA0cE18/vYWquuAd8SvRwFHxK+Pwi7kjor7/0PAQnIL4BVxX03s737Ykf6JPw86ZPfRIU4GgEnAa9gFVZ5s0PuF6oewi6h9gTrgt8DP42OXAU951zwIu/irwuqh5cAH4zF2BHahdXB8boEs9pf8J875F6yurInH8UygLj42F/gHMC1+1k8CX4+PFVuoLgGmx2OjPP5sTuL3JgLL4td5uib+bDZ2A3AYdjz+CHgkcf7D2M3S5PhZzNmZ/ukjmarFzk+3xe9PBA6Nn+EMrI48z5OHZmAWdv74Dna+SNdCNb7Z6bFQrsAqinuxZpc5wALvvNr4oYyNj2/DU/jA+4FHu/mN84AXEh17XfybJ3mfHyOh8T77d+Cn3uB9YiCFoJgwJD7/C/CluA9v8z432J3ZVO+zY4HF8euvAvcA+yWut188ME4FKgajrTvRNx8AVvdwfCFwpvf+NGBJN+ceBmzy3j9GSiaZ+H7CeOm+bwrGCHah+kjis6XAbO/9u9V3bH+hujI+Z3jimj8mXuwm5O64+PUK4LJBlp2gQ7rvm91Ch8Qy0IxdMC7FLnxqkrJB7xeqfwau9L53AHaxUY5dtLcA+8THvgHcEr++CHgycW83kdsw5sliH7Z9ewvVj2AXlIcWOTYX+IL3/pPA7+PXxRaqXyny/TmJzz4K3BS/LrZQvRX4pve+HsvMTvTOPzVxT38cRJnqxOrIgv6Lz/0+8L349VeAO7xjtVhSoc8XqrvshB9F0etRFM2Jomgi1vw0Pm4MwGrvvNb4ZR3Wv6oCWBUHBTRihXwMgDFmjDHmTmPMW8aYLVjT4+jET38MmBdF0aPeZ/sA43XN+LpfxE70wvJdbXMfYwLWdAL597Yn9sE/57XlwfhzsCafBcCfjDGLjDFfAIiiaAHwKayiWhv34/j+b8YuYQMwugffv/FYpSwsjT/DGFNrjLnJGLM0lpUngBFp9c8L42WnkLyHcRTKw4ReXut8rIl/WRwkc0z8+T7AVYm+GJe4bhr6ohiCDtmNdAiW0RoRRdE+URRdGUVR2y5cq1i/lAN7RVHUBNwPXBwfuxjrVgR2vByTGC8fwG6shX4dL8aYskSw1XjsAvlh4NexPvx2QiZWe69bsfq1O/Tm/rfnn5rXv1EUbcG6l3SnV5xcDjDOi6JoBJYt/wTwuDFmrDHmGGPMo8aYdcaYzdh5RHPLeLx7j+esDf1xc30aLRpF0RtYQTlkO6cuxzJEo+MBNyKKovooig6Oj38Lu9OYEUVRPXAplh3w8TFgb2PM9xLXXexdc0QURcOjKDrTv82da13fw9ho7wnYnRrk39t6rG/MwV5bGqI4MjGKoqYoij4bRdG+WLPoZ4wxp8THfhlF0SysMomA/xigJu0s5mN9f87r5vhKbFuEvePPAD6LZQGOiWXlhPhzyUtqnncSYbz0Gsl7WEWhPLwVv27BLs4Ef+IkiqKnoyg6B7vI/z1wZ3xoOXBdoi9qoyj6dQ/3MegIOsRht9QhHnqU+x5QrF86sSZesL7h7zfGHItlbrXRXQ48nhgvdVEUfdy7Vr/2WxRFXZEXbBVF0cooitqjKLo2iqLpWJP0+dgF9E79RE/vjTFVWNeGh7s5HxL9a4wZjjXzv+Wd40fY+3I54Ij79LdY1ncW1t/3XmBSFEUNWFchjYtVWGYYAGNMDdaNps+xq1H/BxpjPmuMmRi/n4Q1Sf6lp+9FUbQK+BPwXWNMfRxVN9UY8874lOHEVLQxZgI2iCGJJqxf0gnGmG/Hnz0DbDHGXGWMqYl3XIeYlKX/idt8FnaSvD2KoleS50RRlMWaI79njBFzNsEYc1r8+ixjzH7GGIN1dO4CuowxBxhjTo4H0VbsRNU1MC3bOURRtBlrRvh/xpjzYoajwhhzhjHmP7HK8mpjzJ7GmNHxubfHXx+ObWOjMWYUcE3i8muw/leDjjBe+gx3AF8xxow2xuyJdeKXPLwEzDDGHBorTicPcRsvMcbUR1HUge0TjY3/Af7FGHOUsagzxpxtjBk2cM3qPYIOycfuokN6wIvAxXGbZwLv6+X37gA+bYyZYmyaq28Cv4qiqDM+/gfsQuur8efZ+PPfA/sbY/4p/s2KeOxM77sm7ThiuT3E2Cj9LVg3hr6S3aQcvBN4PoqiFrCLPCyj6J9zB/BhY8yMeDx9C+syscI75/PGmBHGmL2xpv9f9dH97jBi3XcudjH9OnZsbIyiaKsx5mjgEu/0u4CzjTHvMMZUYt3LkgRJn2BXGdUmrJ/b08aYFuyE+yp2h7o9XIZ1wP0blgq/C2tqA9vgI7DO2PdjHbwLEEVRIzaI5AxjzNdiQTkb62O0GMso/AQbEZoG3GeMacLuRr+EjVb+YA/nX4U1zf3FWJPUw9idP1iH8IexC5T5wI+iKHoMS91/G9v21Vjm6It93pI+RhRF1wOfwUagrsP20SewjvFfB/6Kja58BXg+/gys2bwG296/YE2bPn4AvM/Y/HA/7OdmbA9hvPQNrsMuSF/BysTT2AmAKIr+hp1sH8NGIj+R+O7lwNJ4PH0Y+Kf4e09jA5JuwPbv37HMdNoQdEg32E10SHf4Mjb7xybs+PhlL793C/Bz7DhZjN2Y/KsORlG0DatPTvWvGbsFzMa6A6zEysl/YGVnMDEee79bsIFmD2MXi32B72PZ5UZjzPUUN/tfA/wyPuc9URQ9iF3k/w7LQO5NIcN7H3aj8UJ83s/66H53BPcZW/BgC9YX+fIoil4DrgS+GuucrwDOwhQf/1fsZnkVdn5bSz+XJ7sAACAASURBVD+kW1REcUBAQEBAQEBAQC9gjPk7NoPK33fy++VYxndKFEVL+vLeBgMxI98ITIuiaHFfXjtUNAkICAgICAgI6CWMMdXAzTu7SB0qiN2jamMXqe9grRVL+vp3wkI1ICAgICAgIKCXiKJoaxRFaQ8wHAicS66IxjRsIZk+N9MH039AQEBAQEBAQEAqERjVgICAgICAgICAVKK75Mi9RanTsX2dSmGX+6OtzeZuvuuuuwB45JFHmDJlCgBr164FYN26dYwbZwO+DzjABvCee+65AIwfv0u5glPXH+vXrwfg0Udt+r5FixZRWVkJwNKlNo/yhAkTeNe73gXAwQfb1KIVFRW5m4itBjYLzw4hdf0xyAj9kY/U98ftt9/O6aefDsDo0TZPd0tLC7/73e8AeOc7bYazSZMmFb/AjiG1/dHR0QHAzTff7HREU1MTALNmzaK+vr77mwj6o68w4P3R1dVFJmP5uGLPr7GxEYDPfc5m9Js5cyaXXGIzMEk+xo8fzw9/aJM9LFiwAIDvfc+moy4r26WaEP2RymlIykhgVAMCAgICAgICAlKJXfVRHZKr913ATveHdvxHHnkkAKeeeioAnZ2dvPDCCwBs2GCrk40YMYKzzjoLyDGOb71lC13ccsstDBu20znKB6U/slmbQ1o732XLlnHaaacB8MYbbwDQ0GBTe1ZUVLg2jxo1CoDW1la2bt2ad82LL7ZV/+64I5dCbyeYkdTIR0qQmv649tprAfjmN78JwNSpUwHLkOg5Nzc3A3DRRRfx4x//GMjJxYMP2lSZq1evprbWL+izQ0hNfyQxe/ZsABYvXkxnp83dLktEJpNxDKIYoXnz5vXFz6auP/7yF1tLQ+2bO3cu69atA6C83BoUL730Ui691KbMbWlpAXK6BXJ6Qyg1/TF37lzuueceAH77W5tiedq0aQAcddRRTrdWV1cD1nL3xBM2/bB08/veZ+sHnHHGGe67O4EB6w//mel5iSF95ZVX2LjRVhwePnx43rGbb76Zri5bH2DCBFvldP78+bz00ksA/M///A8AxxxjKy8vW7aMESNGAHD44YcD7Mj8GxjVQhTtk7BQ7Vvscn984hOfAHCT50UXXeTMdBpAq1ev5vLLLwfg/vvvB3JuAbfeeuuu/Hwq+mP8+PF8+MMfBnAuDldddRUAdXW50sxSQG1tbW5h+8tf2pzUmoCXL1/OxIm2yltyQdwLpKI/UoTU9Mc73vEOAF5//XUgt5ExxtDa2grkFhurVq1yC5A997Rl7rdtszmpn332Wfbdd6eLDqWmP4Tly23pbS1Uq6qq3ETqy/1ee+0F5Cboc845B4CPfOQju/LzqeiPRYsW8ec//xmABx54AMC5P6xevdqZb9UfF198sdMRb775JgBHH300YF0iSm2h+vOf/xyAn/3sZwBs3LjRtaGqyubjly7s7Ox0Gxihra3NnafFvIgAYwxvf/vbAfjRj360o/c/KP3x/PPPA/Daa68BMHLkSNdm9Yv0x+jRo5k/fz6Q0y3ZbJYPfehDQG7+efXVVwHbPyKQpFNOPPFEJ0/bQVioFiKY/gMCAgICAgICAkoHuxpMFdDHaG9vB2D//fcHrOlOO/inn34agAMPPNDtiLUT/Mc//jHQt9pvOPTQQ3nooYeAHDOknW8URW43LHeJtrY2128y4Sk4ZN68eVx44YVAjkGJomhnAiNKBtlstoA1/uQnPwngggJKHWqfGA6ZLbPZrJMPBSbW1dU5Zl5y9Pe/2zzd69ev3xVGNXWQSVNBIlEUOeuM+qqzs9Oxzps3bwZ2OQgzVbjrrrvYZ599ADj++OOBnIXlhBNO4PHHHwdyrOnkyZMdEy3GXczqmDFjHLtYCqkcn3vuOf7jP2x6T5m1R44c6XRl0v0pk8m4uUSoqakp0B81NTXue88++yyQY99lDk8rbrnlFgAOO+wwwOoFsZ8Kul22bBlgXejkRqSgu/r6elatWgXk9IbQ3t7u+kas89133+0sowF9g8CoBgQEBAQEBAQEpBKBUU0ZtPMXK/S3v/2NhQsXArkdciaT4a9//SuA8z/z0zGVKmbMmAFYn0LtTsUiizFrbW0tutuXH6/6TezRRRddxIsvvgjkAm+GOqPqMz/PPfcckAuiOPDAA7nyyiuBnM/zLqZYGRQooE6BQmKM2tvbXbskM5lMhk2bNgEUBE4tX77cMWtDAW9729sAHAN0xhlnON+8ZFomgCeffHKA77D/sHLlSsDKsxhlWVokEyNGjOCRRx4BcH7tHR0djhWTjl2zZg1gGbZSYtx/8IMfuNca1y0tLU5nyudU4wYo8NfMZrOOZZWe1LHy8nKX5uyVV14BYOHChY6FTBveeOMNd79q++bNm91rPW/fIqOxI51SVlbm5EN9JbnSd3QeWGuGAvbE0AfsGgKjGhAQEBAQEBAQkEoERjVlkG+U/KLefPNNxw4cdNBBgPWBERugXZyY1VKEUkjJz3bPPfd0DLHvV6f/YgcUyV1eXl7ggygmYMyYMY5BEXYg6r8k4bPF8tNUf/74xz92qc/kB12KkG+lZECMiJgSyLFoURQVHE+msBqqOO2001xqHfmhVldXu/EylCBZGDFihPM5VIohjfmNGze6qHj5sba3t7t0XZIPMfALFy50jGopWGEWLVqUlw0FLAuoz3wmFWx7NYeIQfShceInzde40rVee+211DKq8+bNc/e+ZcsWwMqE2pxMa1hVVeVkQN/LZrOurcm+qq6udteVf3hZWRl/+9vfgFxBjYHGzTff7DLnFIMYYf33290Xcq7+WrRoEdDzXPOe97yHj370o0DOypFEahaqPeW4LBYcIvzmN7/hve99b8G1SkGpFINy1GnxFkWRa7tcAMrKypypWwvUr371qwN9q30GTRwyw1RUVBQoTT9divpDi4/KykqnPPU9nV9WVuYmLZmLZQoaaigW7KE+Uv9s2rSJWbNmAfD+978fyDcXlgqS5jaNd78SjZ+OLJmaTOcnJ6qhhsbGRjcW1PbNmzfn5QmFXaq+lBrIXJ/JZJypVjrzzDPPBGDJkiVuw6+FRXV1dUGuTQWb7bHHHu76pdBHGzZscO4tWqiWlZUVbND8YCqfBIBc4BQU6tO6ujpHnGhMpTmQ969//StHHHEEkJOFefPmuXzKxVzm1A9qXzabdYSJ9I0fjKUcvVqsNzQ0sGTJEmDwFqpXXHGF0/nFUs7NmTMHwKXRqq+vdwtuwXcNU3uTQXk+1DebN292r+V29K53vcu53glKpXnfffe5gOfuMLSppYCAgICAgICAgJJFahjVYrvUpNnB/0y09uuvv863v/1tAJc2o6cdr5+KI40m4E996lNAbrdRX19fkD6kq6vL7ZCVbHjy5MkDd5N9jMWLFwO5HVk2m3VMoHa3/g6u2HPTZ8WOybyrKjWq6jXUILn35V+parQjHjVqlGMRJGO//vWvHSviF1QA+yyKXXewkRwTxdguneNbJZIohZRDu4Lf/va3LqBDpvGKigpefvllYKeKYKQWYvpqamrca7GswtixY12A5rHHHus+l9yojxQMM23aNMe6SyelGZs3b3aWKY35rq4uV+BBbfEDKPXs9Vl7e7t7rWMKIjLGONZZc1CaGdWNGzc6t44xY8YA8JOf/MQFFooFVTvb2toKCiC0t7e7vpQMSE9mMhnHvvtBnLLeDRY+//nPu3H/nve8B4CTTz4ZsPOtxofPmoo1T+r5zs5O137pC33P/8xnotWHcje6//77nbzMnTsXgJNOOgmw88/29HDpa6eAgICAgICAgIAhidQwqsVQjCW57LLLgFx5xL333tulavrsZz8LwH/+5392m3In7czB9OnTgZzv6bZt2xwDpnv3dzPaucjvsBShZNsjR44E7M4s6djt+yImGb5MJuNkJbkb9oOvVG52qDKq/niRM79K/YkBaGtrc32p3e6mTZscC/OnP/0JsD5FkN7xkgwKUdt9/2bpiDVr1ji/u6RcJa8z1LB48WLHGq1evRqwuuWtt94CcKnb5MdXypCPXVlZmWO+5K+pY3vttZfTrfJZlM8q5ORDjPPxxx/vfNxLIfhQfqmQY7rWr1/vxoLmDl+HFrPYJVPXSa9u2LDBBeCIqVRasDRBz2/KlCkFKcpqa2sd46k5R/oviqKCdYffH7qW+mzkyJHOYqe+r66udnL0xhtvADYt4EBAxSzKyso4//zzAfja174GwG233QbkF/fQ8/X9U5Nrp66uroJYAChkUsW6VlVVFcQJTJ482bH56ksVKzr99NP51a9+1WO7SmKhCrnFjJStTDRr1651CkQmnWHDhrkF3yWXXALkJq3x48dzxhlnDMDd7xp8h/jkQzfGOOEoBXNUT8hmswV12Ddv3lxQn7wns3Mxs4FfyUqDTC4GQxV+HylASspA46aystIt5KQ4hg8f7iYdVQK7+eabAVyN67RCE6g/Aet5qzrZsmXL3EJVxyQfQ3WhKn05atSogkjlsrIyp0uUC3MoLFSXLl0K2E2ZgqCUHUJjY8uWLW5hqsV6Z2dnXpYQIE9eVqxYAaR7oeqbmpMbeT8wVQuo5DiA4u5TSQLAz8Gr6yvvaJqgDfqCBQuYOXMmAH/4wx8AG1inNkoHar7151O/kmEy6t93Fdhvv/2AXNWqSZMmOZeTBQsWAAO3UFW+7MMOO8xF3Ov5K1C7o6PDkV9+sG1yIelvYHQN3/TvL+Qht1BtaGhw847cCRYvXuwWwwowe/DBBwFbNVHnd4d00iUBAQEBAQEBAQG7PVLNqPrmB7GlWvUrxVBzc7Pb7WiFP336dJc/TyYArea3bt3KlClTgIHb5ewMtIPJZDKuH/ydbnI3U6oQ0wf5VVCSrMCOBr34gQGCAiSGGooFxChlihz9tXPu6OgoOL+pqcmZvsQOfOMb3wBs2rOLLroIyAVmpQFJx38xPhs3bmTs2LEAHHPMMYBlUpSGJWnW8lPxDCWIUcpkMi6QRn1WW1vrLExiFYcCxIo1NTVx5JFHAoVm6SiK3JgQq2SMcXpCLjIKQNm8ebNjitIMX48mdWWxwCmfLRR8XSvmNWny7uzszMtPDOlM8SZL6uzZs11bJf/btm1zbmBi0KU/stlsQR5V353IvwZYtymlx1RKqmOOOcZ9Jr06UJDe/+QnP+nWTGLbxYavW7fOBV/LbSGKItc2raf0XH359839xVwRwc41yWP77LOPs2hKN0mmXn/99e3KUGBUAwICAgICAgICUolUMqrFAmO0OxDjI0yePNkFjmjn2NXV5XZKYl61U960aVOqazfLt0W+VTU1NW6Ho11KZ2enYwN0nvpHLFKpQL50sOPpj3w/1CRD4F9LO2T/t4YSkv0WRZFLPyK591kQjRPtsGtqalwfiWGU33Bra6vz8UoTtPsXOyY2bdmyZa4t8kX/8pe/nFfH3Eep+3h3BzGJ1dXVBXXKKysrC1IYlTKUuF/P+B3veIeTB0Eykc1mnfxLh/op2NQf8mN9/PHHne+8HyySNojN8/0phZEjRzrGatiwYXnHijGI2Wy2YJyI/ZoxY4YLXtbvSFekEX5xFz+I9vbbbwdyFchkZd26dWvePAv5RWaSqcrWr1/v2Hv9H0zoWUydOtVZxVSFTn6gY8aMKQgmbWlpyavq56O9vb0gTqaYb7/6xGdU1V8VFRUu7kT+46+//jpgZVfrmO4QGNWAgICAgICAgIBUIpWMapIheuaZZ1wUskrhiQ068MAD3UpezGpzc7OLZtXq3fcxSaYwShOUhF2+I8OGDStgCzOZjOsj7XBuuukmoPQY1e58q8R6FEv4nzynWJSq4CcfTmMalb5AkkV+5ZVX3M46Wfqvs7PTjR0dGzZsmPtMLJOYysMPP5wLLrhgIJqxQxAjKPkXa2iMcf6XfiS7GOVkycTtRZuWKsSot7e3O90nRrCurs699tMZlSrExigOoba2tiCrg3RAR0dHgU4xxjg2Sf2heSOKIufjp2NpZFT1vKuqqly7xCqPHz/e+QcqHZPa4qen6knHSrdMnDiRhx9+GMiNuTRmzvCfbVI/tra2un5Ilhfu7OzM8+cHKztqo/SHHzci3emnuEpioIql+ONZWQgky35ci56dn9ZPOkHs+Y7qBj8TQjKGoKWlxfVnMi6gqqrKpcnrDru0UO2v2sfqKDk8L1y4kC996UsAPPTQQ0CuEtPy5cud0Oizjo4OFzgj6jlZ6zit+O///m8gR6P79+u/lnKRMvrpT38KwC233DIg99lXaGpqcvLjC3cyLZXv6J8MCPDNVMm0K2VlZUUDB0oJURQV1LbvCQ888IAbQ5IjbXyMMc4Eo2u2tLQU9KmUymBXWOkOknstUjS+Ozs7namrmF5KfpaseT9UIFNwS0uLM29KJmpra51+1IaklJHchNTX17vFgxZwftWcZDCQrz+0wdPCdtmyZU6e/MDMtEHj2w++1UZ09OjRBdWj/Pmw2DyeTG2l/xs2bCgItEojelqTVFRUOLlQyjEtyvzAKf9avusI5FdRLLZxGawqfnqWK1ascGnD5JLgt0uyrP/Fqvf5ecr910Cem0ByHG3btq2g/WVlZW4hLN0t4mj9+vVOR3WHYPoPCAgICAgICAhIJXaJUe3LXYN2snfeeadjUrVjmzp1qtvBiDXVDnHbtm0uWbl2DFOnTmXDhg1AflCSf07aIOY3abL1UzX5DKJ2MdrNiSVbunQp++yzz8Dd+C5i69atRVnCJOvhy5ofRAV2t5ZkS/3dYdI0tXLlyrzqHKWAnpjU5E74pz/9qUt2L+ZAbJMfGCAWIZvNus/EsEmelLA6bdCuXHKitkRRVMCSGmPcuNe4EhR0OdTgp3tJyn9XV5f7bCgwqjL5K+iprq7OzQ8KnPUZd7+CkL6vOUSQXt1jjz1cAFKa3STEkPuMoB9QVqxoDHSfnip5vs7r6uoqKBpgjHF9Uwrp3nwWWfcrBr6hoSGvMIbOF6Rv/IIQycCzwYTWOytWrHD3LZ2nNVEmk3Hsph9MqbYVS/hfzBqZnHd0Tnt7e0EKxGHDhjl51Gcamy+++OJ2U0cGRjUgICAgICAgICCV6JNgKt+Hrie/MP+YgmjuuusuAJ5//nnAruYPOOAAIJeq6YUXXnA7GK3Glai6o6MjLwUF2J2AdtdKdi1fjMWLFxekmEgD5Hsrf6hiibh9JiDZp0rbdf/993PllVf2+/32FTZt2pSXVgxsm5Lt83d7yWTCmUzGfSZGWs7yUMhGvvrqqyXFqPrjJlmD24d8kbLZrGOB/F0uWMYl6WtXVlbmgoqSaUU0ftIGsQTF0qUoub/Q0NCQF5zoI/l+qEC6raamxgVT+DXaxZgPhfaLRfLTb2nuUPv8EpC+fyFY+e+uHOT06dNdUEqSQUoT5ONXXV3txoJY4ZaWloLAU7XP9zX0GTRfF0NOjzQ0NBR8N4oiJ1ulwKh2dnYWlFSWZWHMmDGufcViIJKlRLtL6TRY8GN3NO4FPwBb9+/rz6Qu9a2aPRXeKWbxTQYxbt261V1P6eR0ry+//PJ2rfN90st+fe3e4IYbbnDR7Vpg+ZVSZPr3rynzjgRF569atcoNUnXU+vXrXQfJpCOauaOjw1Hgql6VBjz11FNA7n5nzZoFwBNPPOHarkpaS5cudQuIo446CrABZwBvvPHGwN10H2DLli0Fi9Jt27blVUKB/Eoqvslf39PCyh8Y+p80965du7bf2tPfSI6ze+65h4svvhjImbH9RXgyUrmzszMvPyBYhaQFvj6Tsk7TZs5HcqHqK9BkTXY/c0bSDD4UFmrFIJ1RVVXl+kqL9crKSjdRD4WsB9J9wujRo/MCx3yUlZUVLMJ8Vyq5yEh/GGNYunQpkFsQy60mTZDpury83D1nkTWNjY0F1RuFzs7OotlT/Ih3yMmT6sVD/mI2rS51xbB161Y37pMbf79/ipn0kybtwQqa6g7Tp08HYP78+W4zmoTv+lMsY0NyMdrV1VVAJvmvNWf4eYkFXb+srMzpIcmSCMn77ruPvffeu8d2pXeLGBAQEBAQEBAQsFujz3jr5O5Du8+33nqLFStWAJYdBGv2P+yww4Acc6P6r376HK3Qt23b5lbv2gmJRR03bhxTp04F4KWXXgLs7lKVJnS+GKOamppUptWQeWnZsmUAnHzyyYBlTMWSqoZ5NpvlbW97G5AzcavKg1wHSgVbtmzJq7kNllXWDi7JCPqMor+jT+ZdFTPtmyEEMQ6DiWRKj2I7+WJmpT/+8Y8AfOITnwCsRUGVo/xckNq1qm99s0zShFlTU1NQa1n9LaYmbRATKBnwn7GYJKG+vr4gTVEy0GCoQXkzfd0p82Ymk3HjKs0pl3qLU045Bcjp+EwmUzCuJN/+MX/OSvaD5OXggw92OjnNgXdqX1VVlUtLJBeYjo4O19ZkQFgURUXnw2TAqvTJfvvt58aerjl+/HjX90k3vDSitbXV6UUxwX6ltqRu9qt3CT6zmqZgqjlz5gDw3e9+160XZEH2c2Un9aA/ryZTkhVL6VaMWfUtdcn+amtrK8jdLH3U0tLCCSec0GO7AqMaEBAQEBAQEBCQSuwSo6qV9LXXXptXNxxyOw6/Eog+GzFihFvBJ3ey5eXl7phW9D7bJKZWu7rDDjvM7SDlP3PIIYe4lb92enq/fv36VKYZkV+hdh2qMNXZ2ekYsNdeew2wu1oxQccddxwAP/7xj4GcL24pQm0v5ofqQ/3h+xUlKw75DvG6lvyat5cKo79QLBVMT+0TWlpaOPHEEwFcnW29P+CAA9zO1GfJ5FumftC4rK6uzqvSo+8laz9rRyxmKW1IBov1hIaGBteOUi36sKNYvnw5YGVHwQvSk3vssYcLNkprQYcdgaxnPpK+h36lqiSblmTgIacjpk+fzmWXXdb3N93H0Dj3/fXVL0uWLOm2P3y21UeSVfQtFrLayde/oqIilXNqdygrK3MyoPv2g66TgUN+8FFSN3d1daXKQjt79mwA/u3f/q0gTZnet7a2utgDycPWrVud3CSDqfzzfFlJFofx/ZSLpTATsyumV7qntraWj3zkIz22KzCqAQEBAQEBAQEBqUSfJPy//PLLXXqpN998E8jttnzfN+1etmzZ4nasWnHrexMmTHBJxrVC7+zsdEns5Xt1yCGHANZnT2yJouJ9H0T52olRKi8vT2Wk69lnnw3A3XffDeAiTQ888EAee+wxINeW+vp6F9ks/1/tdNLYtp6wdOlSx2ioDY2NjW7Hl/R16erqKsoA+KmqIMfU+7tCXWvevHl92YReozcRohs2bOC5554D4MEHHwTgjjvucLtQ7TwXLVoE2FQfyV1+VVWVa7f6QZYF30fVHxtJllXfb25udqnkdA9pgOS8WOq7ZEnD4cOHFzCpOl/tHmrwdWey7j3kCjoMVYY5mR5H8HWHLBEdHR0FLFqp9Ytfcltzrvz+nnrqqQLd4+vFntJuJYuDVFVVOaZWc09NTU3R6PG0ora21ulF9Yv6r7W1tWgi+2J+zZCf1ilNePLJJ93zL8YGJ+Xcz7RTTPaLtTFpFfQzS/gZNXROskiL1nnjx4/frm/zLi1U5WDd0NDAhRdeWPSczZs3O8pX5zc3N7vBlDS7dHR0uAAh33E3WS1CnbJ582bnCK1jFRUVTkHJRK5O8qtopAlvf/vbATj66KMBuOWWWwD49Kc/7cycMu+0trY6R/kvf/nLQE4pffaznx24m+4DtLa2upq/fu5OtTWZOsUPjkrW+NZx//y2trYC87Xytw0Wfve733HDDTcAOfOHxoivEDR4J0+e7AI5/va3vwG4HHlNTU0FuVKLKU+d4ysrX0lrEa9x5isr9V+aFqrJzY2P5EK1rq6uwISZ1rRbfQU/n6UW9QpAraurc2NtqKbnSrqy+MEjeq15oKurq2gFPCgefJhG+Bt6vVb6HyhcmBTb4AnF8ljre0uWLHGk0SOPPAJYHb0j6SkHGy0tLc6FUOmcNB6KVTn00zMlTeCZTCaVAYkNDQ1uDaEAK80ZxXJpd3R09CjnPQVMFUv9V8x1T7pG60Dp4D//+c/bbU/6R2BAQEBAQEBAQMBuiV1iVMVSLl261JmqtUpWeqCRI0c6BstfscuEryAsMaCZTMZ95id7T67o/WSzyWIAra2t3To4d3V1OWfwY489didb3vdYsmQJkKvQ9d73vhewyaz12Xve8x7AmnK0O/rABz4AwL333gvAAw88wBlnnDFg972r+NOf/sRNN90EwAUXXADAFVdcwT333APgEgEXq5oiZLPZAjOEX/9bTJJMfckk4AMFmfSvueYa57yv9sltxWdwxIqtW7fOudLoM6UzGzVqVF6de7DMajLdlL8T1jjRrtpPyZJ0wM9kMqlLag05diAZ9AGFjGptbW3BeclzhhrUvq6uLsdkSHYqKyudnh6q/SArS5JJ7OzszCt0oWNJprGYC0B3rGua4AeLCdKFxWCMKSiAYIwpGC+65qpVq4qmdEtbhaaeUMxNQfrSDwjy+0PtSxaW8c9LG1Sp8Itf/CIAV199NWDXZnqGPaWU8lNCnn766e67AM8884wz3Wu+koyUl5c7ufErSirNm4Lef/Ob3/S6LYFRDQgICAgICAgISCV2aRukXcW0adPcDkxsqFis9evXu7r1/updKVP0X7v9mpqagnJmfjoJrfr9na/YAX225557umv4LILOSWOd94MOOgjIlX8UYzZt2jTe//73AzlG0E9JJLZVO8FSqLWcxEc/+tG890uXLi0o/+YHS0kG/J2xXktOJBNKyQODx6QK8+fPByxDKkZQ/p8KdqqoqHC7djGe/g41WS54zZo1Bf7bmUzG7YaLJarWMT+FXJJBUX+nVZ6UaizpiwiFKb/Ky8sLWI9SCzrcUfhp+ZIMYjabdTJWakFDvUXSB88PGkkWD+nq6ioatOl/L+3wfbAVsyGsWrXKxX0USz0k+GWqkyVRdWzt2rXO+iO0t7enQGTKvwAAIABJREFUMu6jO3R0dLj1gJ67n1YzqT8qKirceZpX9P20pafyC11I52n9oLXFdddd54oI+UHLyXnHj2u48cYbgZyfqc/cq7+kc4YNG1ZQUKCiosKVe7/tttvy7tm3cnSHPuPr/apC/v+A3iG5wNICZsmSJY6m1yJljz32cBsCLfQVeLO9Cg9pQ7FgBT9vnV8dA7o3s/hVzCA3ePR+e785EDjzzDMBuPPOO11O3GR2g8rKSvdaE2plZWVBpH7yP5Dn8F/MvKn/SbNmJpNx15eyUT9PmjSJF154AcgPzhhsaKGqCcNfUCRNn34uXfWHXC+GKkQONDQ0uJyp+i+dAbnsLEMNMnfrOfuuLcnFZ3t7e0GQTJoWH72Bv9AqVuM9GRxVLPjJP0d6RZ/pmo2NjS7jjtDV1eUCYmfMmLHLbelv+Bt56QrJiz/3SIe2t7e786Qf/e+nKZCspyA5uQLce++9vPLKK0AuIO711193C9Ska1gURXkZMsC2P0kY+e5EcvdUBc3TTz+92yqAvXEbCab/gICAgICAgICAVKJ0PKB3E4gpUi7YtrY2XnzxRQDOP/98AB566CGXnkdmHjkql0IqFR/F7nfixIkuuMxPSwX5qWR6qvSkY8UqdQ2WOU/3MnfuXBcsduuttwI5t4AlS5b06v7UXj84qi/hu6DI+T1NkCkzyZ5OnDixwPRZVVVVkNevu939UIGflkltLpa+TExJKSMZ5NTc3FzgDuObxpMpzfz3pc6odnR0uNSFwnPPPedcjeTy4rdP/eab/tWnybR1TzzxBDfffDOQMwNns1l3/TQi6RZWVlbm+kEsoNris+t+BTMxqNIbkpPhw4enKiCxt4Fdhx56aN7/tKO0VjUBAQEBAQEBAQG7DQKjmjLMnDkTwDk7d3R0cNhhhwE537IDDjjABQlpZ3zaaacN9K32G/z648WYUj+VmZCsRCPH8NWrVzt/XjFtaQiQOPfcc/P++5C/l/wMV61axcKFC4HiPkjyH/Orr4kNUH/4u/7krtsPVkw6yI8ePZoJEybsXCP7EbIoiPFRMEd7e3tBII3/WTLdzlDHsGHDnFyINaqtrS2o7FXKSDKqnZ2dTn6TBSz8Sl1Cc3Oz66Okr3up9I/viy6rnDB//nxXUESxIxov2Wy2oD98C40YRI0fP5BK+rS1tbXAipFmrFu3zlkS9Jz9SlXSlTpn+PDhzsKp89TetWvXFswvAX2PwKgGBAQEBAQEBASkEoFRTRkUVagdfnV1tYuyFIOYyWTceX7hg1JFsuzptGnTnI+qSt1pJ9tdGpRk9Lx2t6ecckrBTjdNUZrFoPRpaUyjlhboGarQiKwOL774YoH/aX19vSusILZIpROHKvyMDtIVGhsbN250FplZs2YNzg32IYqVPVXWFLVTfaDjPqqrq/MybEAuzaJfPjPNEPvX2NhYoCMV7d1f2LhxoyvtnExdlQYk4yBGjBjhfDNVFlp91tLS4jIA6HuLFi1iv/32A3J6RxaJcePGDflyzGmA2UUz6ODbUHcNfW3X2eX++P3vfw/Agw8+CFizlJStFmvDhg1z5h1NPieffDIAl1566a78fOr649VXXwXgL3/5C2AXJjLr+ylC9PqII44AYPbs2YU3s+OVZVLXH4OM1PRHUm8Nkok2Nf2RhBZq//7v/+7Mvcq5fOqppzpXIU3AfRRclpr+mDt3rr1AYsz7demlTysrKwtcZPS/WDDmDmDA+uPZZ58F4P7773fuY2eddZb9UhQVpPcrFojaG/iLPsnT6tWrXTXE7VwrNfLRE5TGTanNVqxYURCg1kfoD6U1JOeYYPoPCAgICAgICAhIJXaVUQ0ICAgICAgICAjoFwRGNSAgICAgICAgIJUIC9WAgICAgICAgIBUIixUAwICAgICAgICUomwUA0ICAgICAgICEglwkI1ICAgICAgICAglQgL1YCAgICAgICAgFQiLFQDAgICAgICAgJSiZJdqBpjImPMfr04b3J8bunWGO0FSrE/errn3ranyPfmGGPm7vrdBQxllOJ4CRg4lKp8BJ3aPYwxc40xc7o5tq8xpnmAbymgl+jzhaoxZpYxZp4xZrMxZqMx5iljzFF9/Tulgt2hP4wxjxljNhljqgb7XvoLxpgTjTEr+uA6zd5f1hjT5r3/QF/cayljdxgvOwtjzJJYXpqMMY1xP33MGFOyhMOOYneRj6BT3TkDoi+jKFoURVGPdYS7W+gaY04wxjxhjCmPNwOT++q++hKlrD/69AaNMfXA74H/AkYBE4DrgG19+Tulgt2hP+JBeTy2xvA5g3ozJYAoiur0BywDzvY++0Xy/DQwNQN1D7vDeOkDnB1F0XBgH+DbwFXAzcVONMaUDeSN9Td2F/kIOjWHHdWX/QFjTGY7i7kzgT8MxL30AUpTf0RR1Gd/wEygsZtjU4FHgA3AeuAXwAjv+BLg34CXgc3Ar4Bq7/jngFXASuBD2EG8X3zs3cALwBZgOXCt973J8bnlfdnW0B/uel8BngKuB36fOPYz4P8B9wNNwNPAVO+4f8+z4ns9qcixKuA7WEW1BrgRqOnmfubE9/Nfcb+9AZziHR8P3AtsBBYA/+wdqwK+H/fpyvh1FTAMaAOyQHP8N74P+m4JcGris6/Hz/qOuM/mANXAD+Pn/Vbc15Xx+VcAj3nfL4/7bnL8/izg9fhaK4BPe+eeA7wENAJzgUO8YytiGXsFaA/jZeD1Ry/l5ehYLg/BjrcbsJNmC3BqT2MHGI1d+DXG4+FJIBMfuyqWtSbgTbwxNIjt3y3kg6BTey3/Rc6pBX4Zy0Ej8AwwOj42F7uxmRf33YPAqPjYfkDkXWcu8DVgfnyfvwK6gK3xvX7fO/dlYEZ83Qg79pqB98bHPxb3ywbgbmBc/Ll09b8Ci7Fy+23iMRj0h3effdwR9fHDuBU4AxjpHdsPeFfc8D2BJxIPe0ksVOOxu+XXgY/Fx06PO+mQWMB/Sf6gOxE4FMsQz4jPPa+/FEnoj7w2LgCuBI4EOoC9vGM/iwX4aOyg/AVwp3c8ivvhNKxCPTp5LH79fawiHAUMB+4DvtXN/cwBOoFPAxXARVjlKoX0OPAj7OLvMGAd8SACvgr8BRgTP5N5wNe8Pl3Rx/KxhOIL1Xbg7Pj51QDfjO9lz/jengauic/f3kJ1HfCO+PUo4Ij49VGxXBwFlGEn54XkFsArgOeAiXQzgYXxMuD6pEBe4s+XAR/HjrfNwHFxW6rpYewA38JOPBXx3/GAAQ7AjsfxXh9MHah27u7yQdCpOyT/iXP+BbsYrMHqtZlAXXxsLvAPYBp2Qfsk8HVPfiLvOnPj35set7k8/mxO4vcmAsvi13m6N/5sNrA27pfquJ8eSZz/MDAylqUFyd8I+qOPF6rxTU2PG7wiFu578Qaad955wAuJTrzUe/+fwI3x61uAb3vH9scbdEWu/X3ge14nDdpEM5T7A7tj7yC3Y32DfMbuZ8BPvPdnAm947yPg34GlwKGJa0vhGuzuzmcNjgUWd3NPc7A7d+N99gzwT8Ak7K54uHfsW8DP4tcLgTO9Y6cBS+LXJzJwC9VHEp8tBWZ7798NLIhfb2+hujI+Z3jimj8mXux6ny0EjotfrwAuC+OlJBaqfwG+FPfbbd7nPY4d7CLinmQ/xONuLZZRqRjsdu9O8kHQqTss/4lzPoJdUB5a5Nhc4Ave+08SM9YUX6h+pcj35yQ++yhwU/y62EL1VuCb3vv6uL8meuefmrinP/bT2Cnaf5SA/uhzJ9ooil6PomhOFEUTsTvU8cD3jTFjjDF3GmPeMsZsAW7HUsc+VnuvWwE5N4/HrtCFpf6XjDHHGGMeNcasM8ZsxlLtyWsPCoZ4f1wO/CmKovXx+1/Gn/norg3Cp4BfR1H0Sje/sSd29/tc7ADeiDXZ7NnDfb0VxSMmxlJsn40HNkZR1JQ4NiF+PZ78vtT3BhrLE+/HUXhfE+gdzsea+JfFARrHxJ/vA1ylPo37dVziusn76HcM8fHSX5iAZdkgv53bGzv/F8vg/MkYs8gY8wWAKIoWYMfltcDauN8HYxwUYDeQj6BTewljTFki2Go8drH1MPDrWBa+nfCx317f+eiN/tuef2pe+6Mo2gJsons9OxhzTur1R79Ge0VR9AZWcA7B7rIiYEYURfXApdgVe2+wCrtzE/ZOHP8ldmc9KYqiBiwd3dtrDxiGUn8YY2qAC4F3GmNWG2NWY01DbzPGvG0HLnUBcJ4x5lPdHF+P9RE6OIqiEfFfQ9RzhOYEY4zf3r3J+UiNMsYMTxx7K369EruAS34P7LMaKCR/axWF96V7bsEqE2Fs3oWi6Okois7Bmt5+D9wZH1oOXOf16YgoimqjKPp1D/cxoBhK46W/EEe8T8CyPZD/zHocO1EUNUVR9NkoivbFupp8xhhzSnzsl1EUzcLKXQT8xwA1qdcYavIRdOqOIYqirsgLtoqiaGUURe1RFF0bRdF0LDt9PrCz2QGS95f3Ps7IcBx2YVzsfEi0P+6nkeT6BwplbyUDhFLRH30d9X+gMeazxpiJ8ftJwPux1PJwrINxozFmAtZ5vbf4NTDHGHOQMaYWuCZxfDh2V7fVGHM0cMmutqUvMMT74zysCeMgrP/NYViz3JPAZTtwnZXAKcAnjTFXJg9GUZTFmqm/Z4wZA2CMmWCMOa2Ha46Jr1dhjLkgvq8/RFG0HOsj9S1jTLUxZgbwYayfF9gApquNMXsaY0Zjgxpuj4+tAfYwxjTsQNv6CncAXzHGjDbG7Al82buvl4AZxphD44nOyYIxpsYYc4kxpj6Kog6sY3tXfPh/gH8xxhxlLOqMMWcbY4YNXLPyMcTHS5/CGFNvjDkLu/G4vRh7tr2xY4w5yxizX7wA2YKVjS7z/9s79+goyvv/v3c3CQmBJEAQSOSiXLzgBRGvVWu1alFqPS0eq63itWpRK0cFj1ovv/pt1YrWS9XKaau1oohWEbFVkIogqCheAJUAQQjBICQQNskmu9md3x/D+zPPzg4hgd3NhH5e53CWbGY3M88881zen1sgcFAgEDht50TcDHuyiru/P9v8D/QPHVP3kp399rCAHaW/A7YbRbr67mYABxo/fx/AMsuyGgF74Qzbh9o85gUAVwQCgSN2Pk9/ALDQsiwzLdfkQCBQEggEBsE2/c9I0/nukq42fqRbUQ0DOA7Ah4FAoBH2ALICwE2wo+1Gw3bWnQPgX+39Usuy/g3bL2g+bKl5vuuQXwP4f4FAIAz7QXgJ/mBfbo8JAP5uWdYGy7Jq+A/A4wB+EehASiPLsjbAHlinBAKBKz0OmQL7Oj8I2Ga9ebAdtnfFh7Ad5rcC+D8A4y3Lqt35uwth+5VtAvAqbD/NuTt/dy+Aj2FHcS4HsGzne1RvXgBQGbDNINk0z9wDe0G6fOe5fQh7wINlWV/CDrZ6F3Z05Xuuz04AsH5nu10B268MlmV9CNuB/knYpqgK2CpUZ7IvPy/pYvbO86yC7Vf2EIDL2ji+rWdn+M6fG2BHNz9hWda7sAOS7oP9/NTAXqTclvYr6Tj7ev/QMXXvKYN973cAWLnzul5Iw/cCdh+5cOe5PgRvs/9dAKbvPOanlmX9B7Yv56uwlftBSFV4ZwP4DHZmiVdhWwkyRZccPwLJbieKoiiKoihKWwQCgQoA4yzLqtjDz+fAVnwPsCzrm3Se276G7ysSKIqiKIqi+IVAIJAP4K97ukhVOoYqqoqiKIqiKFlEFdX2owtVRVEURVEUxZeo6V9RFEVRFEXxJe2OItwFXV2OTXcuPG2PZLQ9ktmj9njxxReRn58PAMjLywMAJBKJlOOCwaC80lLSrVu3pN81NzfjRz/60Z6cBuCT9vARndoedXV2ju4tW7Zg8eLFAICGhgYAwPXXX9/mZ++8804AwNixYwEAkUgEADBq1Cj07t27I6dhov0jGW2PZDq1PdjHm5qasGiRnTa0rMxONHDMMce06ztqa+1EB8uX29mchg4dipwcexk1YMCAjpwOkJlczXvcR3htq1evBgC8+uqrAIDLL78cBx2UnBBi5syZ+PjjjwEAV199NQDgwAMPRBrwbBNVVBVFURRFURRfsrc+qrrDS0bbIxltj2Q61B4bNmwAANx9990oLbUrNpqqKeH/AzsLx1iWJf+nopqbmwvAVtxuvNEuWNOnT5+Onr/2j2Q6pT3uvfdeAEA8bufQLi8vRygUAgBMmzYNAHDkkXYho7Fjx4pCWlBQAACYNGkSfv7znwMATj/9dADAp59+Kt9/8MEHA7DV1Q6i/SMZbY9kst4eDQ0NWLduHQDIM9KrVy/EYjEAzvNCZfXEE0/En//8ZwBAOGxXhR0xYgTKy+2Kp1QPV61aBQDo378/Nm2yC0k1NzcDsJ/Hvn3bqkYr+EZRvemmm7BixQoAzryzdetWeaWi2r27XQTRsixUV9vFtY47zq7KTYV1wYIFGDFiBADH8mfOV7vBs0321vSvKFmHm6tAILVPf/TRRwCA+vp6ALapvEcPuzLgwIF2pbr99tuvze/2+t7OgANG37595dxp+uciJTc3VwZILkoByEDMxSh/3rZtG7Zs2ZL0O8W/8D5zkl21apVMCGPGjAEA7L///mhtbQUA3HDDDQBsdxEAWLx4MQ499FAAwFNPPQXAnlyvvNLOAc/nhYvTeDyOmhq7HDpf+/dPqsqrKF2GTZs2obDQLrbXs6dd5TUej8vYd9lldq77++67D4C9YVu7di0A20WAx9PN5rvvvgMAmVPC4TBKSkoAANu3bwcAVFVVtXeh2ulw7pg+fTqKi+0CYVxc8rnPzc3FRRfZxdgWLFgAAFi3bp2IJ1VVVUnfed111+Htt98G0KEFapuo6V9RFEVRFEXxJaqoKl2KeDwu6hKZP38+/vUvu2Lijh07ADhmzsGDB0uACXe8RUVFOOSQQwAAl1xil9CmiuoXNRVwzPUFBQXyf6rJZhvQmd/tAgA4SiqPycnJEbX5f5m2VHkA+OqrrwAAr7/+OgBgypQp2TkxF+6+vnDhQgna+PLLLwEABx10kPTx/fffH4ATJLVmzRoJIhk9ejQA4Ne//rWYLvn90WgUgP188dlhwEjfvn3lOLfCqyh+hGN9U1OTqKfs48FgUAKH+Lw8/fTTAIC1a9fKZ8nw4cNRVFQEwOn/VFYTiYSMsVRuW1tb5TuotvqVefPmAbCVYQbsuueRrVu34rDDDgPgmPfj8bhY8KjK8vPbtm1L+3mqoqooiqIoiqL4ElVUlS6FqeTMnDkTgJ1Gg76bgwYNAuA4x9fU1IivEXeKO3bswBtvvAEAeOuttwA46UkmTZqU6UtoN1TCCgsLxb+K7/FaotGo7HzZNrFYTBTYlpaWpO8MhUKiAHRldqeI7g6vwDOybt068fWkUklftrb8m9OJW7mkj9zixYsxcuRIAMA//vEPAMCQIUPED5XHn3zyyQDs82cfZ4BVJBKR72OgFf9eNBqVPkYFasOGDTjggAMycp1dkcmTJ4slhkqTKs3+orGxEYAd/MOxguNe9+7d5Zmnssr7NmTIkJR7GI1Gxa/fPe6Yx9K3s6CgQJ4vvyuqH374IQC7bdwpDzkODB06FBMnTgTgxEgUFhZKn6e1jopqdXU1vvnmGwB2e6YDVVQVRVEURVEUX+JLRZUrdXOX2tGd6mOPPQbAifS79NJLAdi7nnRFoimdy5IlSwDY0YmMWKQC9sknnwCwIxIZoclddq9evSSKnnz99dcA7OTpfonYNDMXMKrbrZSa6eXcKan4WcDZ7efl5YlPUVcmXb7E5vfMmjULADB16lQZI9huLJKwbNmytPzd3eEe7xid37NnT0kp9cADDwAA5s6di6OOOgqAE41MpfS4447DK6+8AgC49tprU76b/YPqaV5enigpZM2aNaKo7uuKoZdS/+677wIAHnroIQC2XyPTFZF9bU7pqMXCfTwj559++mncf//9GTjD9hEKheQZ5thpWpyIqbBy3cHPhUIhOd69NgkGg/I7jqvxeNxXsQ5twXkyGAzKnMLr4ZzTvXt3WUfxvUgkIn673377LQBnrRWLxWRuTpei6suFKgdDr0GRDcXfeXWIyspKPPnkkwCcAeTcc88FYA/gaqbZN+DkGQ6HxXmd5h3mwCspKZFFBs399fX1srDt168fAMjPdIj3A+ynBQUFsvhkfzcHXw4wvPZQKCTmGPM9wF7EclGi2Fx88cUAnHRgvXv3lqA8vl544YWdc3I7ef755wEA3//+91PM9rW1tRIcxTRTTC1VXFyMW2+9FQBkA1ZXVyf93e1aUFxcLItX0tLSIm1D15p9Ffd8MmvWLDz66KMAnOfxiSeekN+75xI/pbdrL16LUv6fYwv706BBgzyvz/3e0KFDAQBz5szBT3/6UwBOvs1sYI57XlX83Pk9zTbg2Eksy5LjOM+wXcrKykQAYRsUFBRIv/A7a9asAWCfO+cFd279nJwcuX5zoc7jKQ7R9J9IJLBw4UIA6Rs3961toKIoiqIoirLP4BtFlUppTk6OVJJ47733AAATJkyQ49y7HS8mTJggCgGDIKg+JBIJVVK7MKZiwRRT7733nlTRoPmFAVSnnnoqjj/+eACQFFYFBQWiuPKVShErb/gB7vpzcnJSHNdN1YvPjqnumKYpAEmpRLrKbn9PcSsCbSlc9957r4w3DHwYPHiwmPhpLmdwVTYwU7BRyWLKqP322w8bN24E4Kh4PH8AKYFQ8XhcEnebAVP8P8dJ9qc33nhD3AioGpWUlMjf2JcUVapubjMw4KQnmzNnjlTl+eMf/5hynHsu6WpqKpBqdTGv6ZRTTgHgJHUvLS2VPsna7uXl5RLMR9V03LhxAIDHH38c69evT/pdJmG/NgOnOCdwjhgwYID83j1OBgKBlPHDPI7zA1/j8Tg2b94MwDF99+7dW4K0/G695X0NhUJJLmMAkgrJcPzhmNC9e3dRUt1zZiKRkGCqdKGKqqIoiqIoiuJLOl1R9Upg/tvf/haA4+g7Y8YM8XM56aSTADi+WCZMMbRx40ZJbv2HP/whQ2feeZgBYVTduINpamoSPxr+bu3ateLPefjhhwNwHKCZXqUrE4lEJKkzd4Us79arVy9UVFQAcJL7P/fcc+KbyoAlvwRQmVApDQaDcr+pmJkJ2PkM8X6bPln8HP2oQqFQl1R9OoIZBLEr3nnnHQC2DyKfCX7umWeewc033wwgu0oqMc+bQVRU9crLyyXwjyrHuHHjUFlZCQCiXjGReV1dnXyf2/cUcPoTVdcRI0aIYku1dezYsVi8eDEA20e2K+H2vzQtMm0pqaz3ft555+Hss8/eq7/pd9gvqJaFQiFJW8TE9uwf0WhUCk5wTKmsrJT+MXv2bABOud7q6mr8/e9/z8ZlAHB8yjkPmEFuK1euBGBb0ehDy/5PS9Wu7pk7wIpBi6tWrRKFmQG6tN7ybwH+TVPFdHtVVVUSx8O2eO655wDYBXLcKnMwGJR1BtdanEsbGhqwevXqtJ6nbxaq7FDfffedDLIciCsrK/Hggw8CAF544QUATke57bbbJKrVlODpAE846Zt0tUhNtlVra6sMLm+++SYASGTlkCFDRLKn+S8SicjClAuy6upqAPaD5nYs9zOm6wYn0ng8jl/+8pcAnGugWaKiokImXvariRMnYsSIEQCciiRe5p7Ohve4oaFBBlBeHweTUCgkEwzvd35+vhznrkz1v4B7gWouTj799FMAwE9+8hMAwMiRI6Uf8XdXXHGFbJZJZ5nwuFDg5FddXS0T43//+18A9ljIjRc39Mz7WFxcnFJ9CnCuh9/7+eefA4A8R4ATsXvEEUfI8+R3U6Yb98LD/JmRyZZl4bXXXgMAiWSmKZuvgLOoyc/PT1r4ur+/qyxQCcd/0/R75ZVXArA3+uYx8Xhc5l4GnhYXF4sQMmrUKABO227btk3eywYcCznfFxYWYtOmTQAckSsUColrmNdmxS0CmfB+cwFaVlYmC1RWbTrkkENk7cI+47eFKsc8rhUsy8IZZ5wBwKl4R4LBoBxHM39zc7OMJ9/73vcAOOLXqlWr0j7f+H9loiiKoiiKovxP0ukyi1vF22+//XDfffclvVddXS07AO5oWGUlHA5LbVmqR6eddhqGDRuW9B38nNcOys+YahBfTRMepXuaYVpaWlKcm48//nhRE6mIzJ8/X37fFZRUL6h+bN++Xar0cHdH8vLy5JrpHnHwwQfjmWeeAQAsXboUAHD77bdn4Yw7BlMIbdq0Sf7vVmt69uwpijEDXkaPHi3XzL7CHW4ikfA0Ae/LBAIBfPbZZwCAE044AYAdZAfYCgHN68ceeywAJ1+mCRXE5uZmURfpXpIJqIhSkWGA09dffy0uLFS2br31VlH++OzTdHv44Ycnqe/8Tn6W5jqqp2Yg1x133AHAfkZo7mWaqq5aqWr58uV46aWXADjXMGTIEBlHjzjiCADAF198AQBJOWWZfsekq6mnXrjH/xdffFHSFnGepZnXnFu88oyyP9Fil+3ATXfwaDAYlD7LZyQWi8l82d71gDs9E6+9T58+EkRFS0d9fb08c+6cxH6BFhTTysK+zPmENDc3i0JNZTgYDMp6i+r00UcfDQB49tln5T22CV1H9pSuuUJRFEVRFEVR9nk6XVH1wu2MXl5enlIJhMecccYZSWkUAOB3v/tdyndy5xQOh0WdHTx4cAbO3hsvB3u+Z+46zTQZ7uPpAzNt2jT85S9/AZCqAnkFy4RCIVGBGGzh5X/jF9qbNJu7+5KSEtnxz5s3D4BT63zLli2inFBl/+qrryQtB330TJ8av/jhUTE2E04Ttk84HMZpp50GAPj3v/8NwFYz3Om2uLMPBAKeypDfSGdAyvLlyzF27FgAToUpKorvv/+++NoxfZkJ2+2ee+4BYPuCM1Dk6quv3utz2xUco/hqpqdy+/wNGzZMLAMMGOH1RSKRFGtKKBQSVZ3v8fmViyQOAAAZaElEQVQx+zz9FC+//HLpRxxH+Mq/41fcz/Kbb74pKhLTK23btk2CSt3jYlVVlaTB8+qL9Imktebhhx+WQJ1bbrklnZeSEczAXAbPXHLJJTKvUDVju0QiEVEQ+RqJROQZouLG54ZKdbagosv5fvPmzTLW8z323fZiWZb0H7YD2yUajcrvqNKuX78ew4cPB+AdG+MHqDLz+SguLhbF+c477wTgzD/5+fnSZvTNLSwsFCvfW2+9BcAJtCwqKpJnjMGJqqgqiqIoiqIo+yS+U1TNKEp32h0gVeVatGiRKAZUC1955RXceOONAGw1BXCiu+fOnYsLLrgAgKOSZAMzStSdPqetCLlEIiFZDaiE5ebmSp1vttGzzz4LwFYHuNPlbn/AgAGyw6ECwp/r6uqS0mn4hfYkbTej/pmeirs8+tz17dtX1Ci+mmVSy8rKAGRXXW8v7vrSQKrPaV1dnaQuYoT466+/LhYId/oVP9Wh5nPgVbrRXSoWSC536C6A4JW5Yu7cuQCACy64QIo+8JmjD+jGjRtFcSGrVq3CXXfdBcDxYWYpwBkzZnQ4XdGewL/L+8i+7pVw/8orr5RUflQ+6IMLONfM/m8WhGD/GDly5C7PJZFIyPfyePqeuWMB/IZ7vpgyZYrncUwGT6sVx8QJEyZgxowZAJzk97W1tVKMhr7PVNBGjRrlmTqxM2nLQmU+L/R9PuSQQ2SMZMQ854vi4uKUjAe9evWSsYr9g1ZOzs3ZguO/WfqV58axorm5OSUdFa/FHFtM+B7nYCq35rGcg1avXi3zCediv8H0fFRIS0tLJbUXx0OOPYlEQtqQvqeBQEB89Fm+md8VCoVkLuLa5Qc/+MFena/vFqpeD5Q5sJIVK1YAsE0LNEux8sWOHTsk/RDzebHRc3JycM0112Tm5NvAnJTdZoSVK1di7dq1AJxOQlNSTk6OPBzs/AcffHBS+hnArrAD2Dkg+ZDy+8PhsHwfzXl8MD/55BNJS5Fu9tR8297jOYjGYjGZSGnC48D62GOPyeLu0ksvBQD069cvxTmeA6uf4CDX2tqaVBEEcCaY1tZW6U8MeAGSF+OAs8ANh8O+GTy93FsIJzwurtyY1w8kb/YYjEl3mJNPPlnuMwdQfm78+PHSNtzAfvTRR7j88ssBONWIOAgvXbpUzi2TZu9HHnkEAPCb3/wGgGOaN9Mlkfr6epk4eS10Adh///1l0eU+xoTBM15BUtdffz0ee+wxAE77sT39ulB1zxe7c+PhpMtgO7oQ/exnP5PAS+bbnT9/vgTwnnXWWQCcQJJoNOq7VHC7G0+50eUmaMiQIbJo4bhqLvJo3qdZv3v37mIm5qu5kc4mbleDgQMHJm3aSEfTEbrFJb7SBQBw5peWlhbfu1e5NxDjx4+XzRfh4rSxsVHGYV4X5yPAcbn8z3/+A8B2iWK6tzFjxqTlfNX0ryiKoiiKoviSTtn6tTdYxsR9PNNfxGIxUcO4C3zggQdEUWJ6DRIMBiU5djZwFzQAIKmUuIs/6aSTROWiCY47l9LSUjFL/f73vwdgq6gsfECFjebI1tZWUU6o0p5//vl48cUXATgmcdYwnzZtWsYU1Y7e47YUWK8AJzNlGVUuJkFnpZT8/HwxyVx22WUAbAWBCgh3hlRI3H+jM2E/jcVicn3c5ZpJ3LmjpdWgpaVFlBD2I7Zpa2urqB6djWnedweLcQff0NAg12Ca69wFEKh4XnXVVZKaiQm+m5ubJdCBaiIV1SVLlshxLAIwZcqUlFQtVDR79uyZFbWE6h3HCBbq8ApKuP322+X6vWgrWT/bg25Ty5cvl79N+KyY38Vnyk+Y48eunuGlS5eKqZLjo1nnnKoileJzzz1XgkRuu+02ALabAMdRzkO0cJ1wwgm7tAJ0FolEQhRBPkO0OPXu3VvaiubZr776Sp4PqvG879FoVN6jBeeDDz6QeYvVIWnpYxokPxGLxfZ4jOdaw2sM5ZjhdzUVcNYSfAUc6wDXIhwbzJR87NtmcQgq7HSbmT59etL3pgNVVBVFURRFURRfkhZFtS2FNJFIiB8HV+F7EszhVlzoM9XS0iK7OO5yn3vuOdnh1dbWAnB8SSKRSMYT3FuW5amkArbic/rppwNwUuVMnz5d0oHQt9aEJdqoIDY2Noo/Lp35GTiyfPlyUQ7pc2L6pD3xxBMAnHKRw4cPl9Jnpo9jZ9BWvzB3wIsWLQLgKERDhw7Fu+++C8BJu8H7HQgERGlnn4hGo6KmUHmvqKgAAAm68QP0gRw4cKD4FLE/sz0GDhwofYw725KSElFXqbRxl19YWOgbRZV4PY+0ANxyyy3im837Djg+rNOnTwfgBEv2799fxgOqALW1tTL28HNff/01ANt/leWHqRZUVFSIAsU2pRJVWFiYUlAj3dAiAjjKFJ/9aDSaothFo1HpH3zWqZht3LhRfkeVLC8vL6WsKp/9mpqaFEUVcO4Rg6h4T2pqavY69Uy68Er9x2IOVNnz8/PFF5lt6wXvQSKRED87Kk0LFixISRTPPnT00UeL+p1J3D64banJwWAwpSwog79GjhwpwV9sox49eiT5tANIen5ofWJxlXXr1mHBggUAnHmFyuOBBx4o1ohslhE1fVC9yt3yemhRM9vHazxyl64mbVkyuhqcA9lHOG8CjoLM8TMWi6UE9rLgjEm60gymfaHqzg2ak5OTJBO3B36W35WTkyOmO5rp+PPkyZMlkpMLs0cffTQpAg1wotWyMYi0Ve/5s88+E3MRA6dqamrkJrPGuNcNZi7I559/XqLVaYLjdW7YsEEWtiYMCGAOSA6+iURCFm2dvVAl8XhcJlV331mxYoXku+SC4YsvvpCHZvPmzQCcBWhTU1NK0MuQIUOkhjUXfu76xn5i3rx5sqC44YYbADjZLH71q1/JopyT57fffou7774bACRwkP1+5syZvlmMmwGG7ueFUfdnnHGG9M+ZM2cCsCdDLloZLMn+X1dXJwMnF1D9+/cX0zaDCv/0pz8BsIMIGEjEScsM3nRXiOrfv3/GXUOWLFki95RmVraBVyBUKBSS/szzNiuTcSxku4TDYXmu3BkBvvnmG7lmc6xkkAqP53nU1dV16kLVa0GyfPlyCebghp6blz59+uCvf/0rAODll18GYAeYMniOLkMUABYsWCAVurjA9YLjcTgczlhWjfa4NnixceNGTJ06FQAk/zYDqPr16yfuLWZeWY6j3OzxHkciEVRWVgJwFmm5ubmykGEwK5+lVatWiasF70U2aOseBINBeZ7dmYXM/mQuWN35dc38srsL/OxquDdhkUgkpcJWbm6ujDE8nhtiM5OQOwhtT1HTv6IoiqIoiuJL0qKomjsP7mTMdBZ0sr3uuusA2MFErKnsVoOAZDUAsHf5dPSmGvT888+n/G3TpEkFit/FnzOZmoc702XLlok5hWoeX3v37p1Sm33YsGFisqbyS7XQNEUwz93s2bOl3ZiuZty4cQBsU7A7aCIcDotSwCABs1JXNnCnA/HKj0lCoVDKDmz27NkAbMWU105l2qxXzHYwzRBu8+3AgQMlvRkVA7a7H9m6daucH6tP0RQ8cuTIFHPU1q1bxdxGJZH9afbs2ZLyKBvWhbbwMrFxjOA9GzhwoJibqbIOGDBArpnplNi/vWhsbJSgJNZ5p0Vh5cqVoi7yO7t16yaKE59LKprZoLy8PCWIhefvpUx8/vnnOO+885Le49jpdbxXVTqOGd26dUt6ngircFGRI17HZhMv5ezll1/GxIkTAXhXRqJr1LRp0wAAd999t1SfevLJJwE448esWbNSAse8XN3oetSjRw/5rnRj/k3mzV22bBkAx62juLhYzPCsNDVgwAC551SWqcBXVFSk9JGmpibpP7TO8fP9+vUT5ZBuN62trfK80K2N8+ySJUsy7mrXXngN8Xg8JUUkCQQCbZ6vWz3Nzc0V9bgrKapeFlsGbbtdm8zrcuefBZxxnO2QiTHBHz1IURRFURRFUVzslaLKXXhLS4uswrmz486qsLBQnK+5Cv/0009FUXX7QwCOGkA145hjjsEvfvELAI5vmRessQw4q3q32pTJ1FRUqAoKCvD+++8DcNqDf/ess84SdYzVZ6qqqsTnlon7mSqHDs5A8o6FSiiDr5jge+nSpaK2caeTl5cn94CqM8+rtrY2K/WY3QpEWz4rlmWJ3+DChQuTPl9bWyvtZwYIsf8xPRVV6zFjxkhhAKok9fX1osRRuWOfa2lp6bBPdaaZOnUqbrrpJgBOajP2hTPPPDPl+IsuukiUQ6Yx4zWNGTPGd5VzAODWW28FAHluqEqtXr1a7jPPOycnR/o/7y2tDWwfk7KyMsyZMweAY4Hgs1dUVCTPCQNBwuGwPB+0hFCByoY6VF9fL/2RCo5XRSoG/BQUFEh/5rPvpajymrwKqJhV8rxqofP7GYDk9lnzAzynYcOGtfkM816yDjkAeb7o285+1KdPnxQLlZeKy/tD385MctVVV+Fvf/sbAGf84jwajUZlrGdqxqFDh8q9p6WF59mvXz+531TUWltbZd7k93J8jUQiEijF/pebmyu+30z6bvot+6UAAvtHIpFol8+kl2rIz7E9gc63KuwJXooqY1TYH8xUfu4qgmaQPPsGv7O+vj7tfuuqqCqKoiiKoii+ZK+2OtxdmD4N3D3Rx2r9+vUpPj7XXnstJkyYsMvv5a6P6S/OP//8NpVUQr8Y00fI7VORyQhV7uLNBPq8Fr4ecMABop6eeOKJAOzIYu72TL9SwI5Qp48I2/niiy9uUzFglCV3PDk5ObLr4+e4+9myZYtnSqx0w+vjzpQ/19XVifpJBb2mpkaUAqpoH3zwAQA7IpWpl5hiaMuWLXLNVLXZVvTlApw+WVZWlpL4nX6QDQ0NvlNUa2trJZqWmSJ4fWYydrJjxw5RydlG7AvpKmmXTj766CPxtWNf5PNSV1cnaeeYLsWyLFH2GG3NjB/Dhw8X38NJkyYBsLOA0DePvoRUiszUPXwmBg0aJCnbqKaxv2aj/OzKlSuToqwBp5iDCceFwYMHy3m5nzNTzXJnBDAxI3jN9FhuOC4x8jvbtdyJqQhRxeN5X3TRRSmqcFtpcsaNGycq68MPPwzA8QkGkNI/vL6DSmwmy+ryml577TXxHWWkPseBsrIyeV5oIaioqJA5x63+xWIxuadmph5arfjM0apBn3fAuwwpx2n6rAKp5Zw7CzNLAZ9/XrPpv2qqpUCyFcWd5gvwZ/ntjrJjx46UOA1a7fLy8lIKRcTj8ZQxhm1ZVVWVdP/TwV4tVF955RUAtoM9TUPMz8aB07IsuTh2jiOPPNIzBQpgP1w0ezMVFXN/AqnBV16O7QUFBfJgskOZtWmziZnbMlt0pJNkwwxcWVkpaYRocmI6qO+++04GXS46SkpKJE/j22+/DcCpHFVaWipBVJwUSktLxeRJ9w/+vG3bNnG7YCqqcDgsgwsnKPaPTZs2+a7qTjgclgUmg8UIzXXu47lZ+vGPfwzASeeWjU1Je+HG5JprrpHJzMz1yVfeGx7T2Ngo/YH3ir/bvn27BF5ef/31AOxcqTNmzADgBBNykG1oaJDFq5nOyp1W5Ysvvkj6fCbxSkHlFajBczM3uhxnzByIxExJxTGZf8tckHstZAmr13DS91pApxvLslLuh1cQyDnnnCPvLVmyBICTq9prcXnPPfcAsCffyZMnA0heoBKvPJxeVb4AezzLFHR/i8VisunmOXFMbGlpkXvDzXc0GpWFJjcYPP/GxkZpW86tlmVJX+FzwvHysMMOw+jRowE47eLVtvx7ZWVl4s7T2WOPmZLK7cJjbuLc1+OVssp0BeCz5H6muhLNzc0yzlIE4c/m2olt0a1btyQ3APN3nHvTiZr+FUVRFEVRFF+yV4oqlbvS0tKkBMmAo5YMHjzYU1LmTpd1lBn0UVJSgvHjxwMAHnroIfkMV+1ewVduqqurU4JkGESUyWAqxZuPP/5YlAcmzWbw18aNG8XMSjUjNzdXTEjc5bOvNTU1ibpEU9+WLVtEyaDbBxW3L7/8Uo7je8FgUL7DraBUVFR4VubpTHr27CkJ/s2gL8C5JpOSkhJRrBnUyPbPpOLTUXgtp556qowRdFWg+h2JROSc2RdKS0vleFpm6Aqwbt063H777QCcAJmZM2eKWkoVnsptIpEQpYeqSkNDgyhWVKfYh7JhGWH6pN1Bhai2tjYlEb875RaQrP7xWt2FNeLxeJuKKsfrbLK7BPcMoPzhD38IIFnlourD9FpPPfUUHnzwQQCQFHUTJ05s1zPfVhJ5tnMmq5adeuqpAOwxzV0Agu5TpaWlomyZ5m32FfZ/zqemqwJV0FAolJK2ieNNQ0ODzO20xkUiEelH/Nu8X/n5+W26kmQDnpNZjMitiLeVLnF3BRz4WarOXVFR3b59u9xDXq/bPc58z7KsJNdCE/ZF87v2FlVUFUVRFEVRFF+yV4oqAzyY+NeESkRVVZUoIky1VF1dLTtCrr5vvvlmALYfjlfA067Swnit2N9++23ZAdLZnztK+tIqmYcBTvn5+XK/H3/8cQDOzqypqUl2pFQ5zTKYDAqicrZ69Wrxd6UvTCwWk89SSWF/6datW8qOnoE6QKqi5OWj1tmYKgn7NXfvXhaGwsJC+T3biNfcWcEvXlDNOeecc8QnmemmOC7U19dL6huOKdu3b5frYT/i/Z46dWpSaisguQiGWcMdsNUV+rfSDzUnJ0f86o499lgAjg9sS0tLpwaHmMnKOabV1NRIezFNEp8fMxWVl18lj+Pzs6tAwl35ZGYS3u81a9ZIm5tWEcC+j1T2mNIukUiIBe3OO+8E4Cjur776qgTenX322QCcVIAdwT0fUUnNZLAdgwMnTZqEd955BwDwyCOPAHCCBNkGJjk5OSmql+l3255gIDOgjP6w9IE1/RXdKYs2b94sqSg7C7dFt6CgQM6TmP3aXVa1rZR0wWBQjnd/Z1fCS1E144HYBqYlhse5U3aZfTBdimrGEpzxgT300EMldyFNF5mmsx8MxYaR3GZVH3Z+LqBqa2tlEmGgVW1trSxYGARkZing4swcfPl7d912BksBzoPUq1cvmcj4OR6XrUpdHaGoqCgpIhdIzpvoxpw4SEdcZ7IFz2Xo0KES0cz7MmrUKAD2/WBksxl4yQWnaWLk++wDPL5bt24pkanshz169JA+wMwAPXr0kJyC/NvcMPkpIwQX+o2NjSkLcBIMBlMWmeZ7fCa46G1tbfVcjO5qgWounNMN3T9ef/11WajyOjku5Ofny0KEi7by8nKJTmegJoPuli5divvvvx8A0lpBimLM3LlzceGFF6bte3cFXXr4SpqbmyWQjELB+vXr5flym7pbWlokAJqb9KKiIpm/uYHh8xMMBmXM4XesWbNG/s/P0XWmoKBAstt0Fu6FuNfC0wyOcmczMKtVebnFuBd2XQH3Na5bty4lAwJpbW1NysEM2P2B77nbxCvAd29R07+iKIqiKIriS/xRMkLZJ6GysGbNGjFV0WTLXXlLS4uon9yZBQIBUVkJd/TFxcXyf+6UzWo67vRlffr0EbWNO8CioiJR7rh75M///Oc/cdxxxwFITZ3WWZimKipJbVFYWJiSMoSvbAs/YOb0dTvumymp3PfBDHZyK8fdu3eX47n7z83NFZMxFSL2r0QikVQnHbBNheyfzEXJvpmpOu7txVQveS15eXmiiLLd2LZmuim+Z1ZfM4OoePyenk+6YT7Qu+66K2N/Y09x97s77rijk84kmfz8fEnPxtdMc8opp2Tl7+wpHDtNFdkMpgScfhyLxTwVVeLu74lEok21tauwffv2lByxpmnffd2BQEDGbLeLBHN9m9+xt6iiqiiKoiiKovgSVVSVjDNs2DBJlG76mgJ2OiKmO+FOLBwOpzj/c/eWn58vqiBV0AEDBiQliAeS61ZTeTJrVNMXlYqSqdr5SXUEbKWPCjGVAHd9ZROvOtvc2frt2oDkAEcqmVSOGxoaxAfRLBziTo1i+uryGtk2BQUFKWmleN9bW1tTUvbMmzcvxTeP7Z3J9EPpgNdJZdVU49mHQqFQigpExSQajYpaTTLph6oomYb93+zz7vmFeI2npiroVZmqK/qoupk1a5aMs/TndsdFmO9ZlpWSioxj467adm9QRVVRFEVRFEXxJaqoKhnHrAvM3SejqPmaib9J3OpRJBIRddXcIfL8slHPvaNQYWT7UUH0Si0TjUblfe7y+eqXutu7wq2Mm1kbssWepCvqTEKhkKTTYoYVphAya96b5U75PtVTKrH8nKLsK7gLIJglQd3jfyKRELXU9L3kHEJrjVlK1V2CtivgVo5vvPFGKbzDdIDtjYvg2MG2XLRoUTpPFYAuVJUs0BlmQ6+/SZNEz549fbkYbQsu6Gn650LObaYF7HRPNN9wcKabRKY2BkrncfXVV+Oll14C4Cw0zUArbmq4OK2rq0tKHwQ4QY5HHXWU5GJVlH0B9wLVdH+iCxCPSSQSsuAyxRW32GEuSt25iLsCbvP8mWeeiTPPPBMApAri/PnzAdiucgxGZbBlMBiUtuMGmOMGq46mEzX9K4qiKIqiKL4k4OU8rCiKoiiKoiidjSqqiqIoiqIoii/RhaqiKIqiKIriS3ShqiiKoiiKovgSXagqiqIoiqIovkQXqoqiKIqiKIov0YWqoiiKoiiK4kv+P5akQKA0Xv0+AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 864x345.6 with 40 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "n_rows = 4\n",
    "n_cols = 10\n",
    "plt.figure(figsize=(n_cols * 1.2, n_rows * 1.2))\n",
    "for row in range(n_rows):\n",
    "    for col in range(n_cols):\n",
    "        index = n_cols * row + col\n",
    "        plt.subplot(n_rows, n_cols, index + 1)\n",
    "        plt.imshow(X_train[index], cmap=\"binary\", interpolation=\"nearest\")\n",
    "        plt.axis('off')\n",
    "        plt.title(class_names[y_train[index]], fontsize=12)\n",
    "plt.subplots_adjust(wspace=0.2, hspace=0.5)\n",
    "save_fig('fashion_mnist_plot', tight_layout=False)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.models.Sequential()\n",
    "model.add(keras.layers.Flatten(input_shape=[28, 28]))\n",
    "model.add(keras.layers.Dense(300, activation=\"relu\"))\n",
    "model.add(keras.layers.Dense(100, activation=\"relu\"))\n",
    "model.add(keras.layers.Dense(10, activation=\"softmax\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "keras.backend.clear_session()\n",
    "np.random.seed(42)\n",
    "tf.random.set_seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.models.Sequential([\n",
    "    keras.layers.Flatten(input_shape=[28, 28]),\n",
    "    keras.layers.Dense(300, activation=\"relu\"),\n",
    "    keras.layers.Dense(100, activation=\"relu\"),\n",
    "    keras.layers.Dense(10, activation=\"softmax\")\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<tensorflow.python.keras.layers.core.Flatten at 0x1c60ebfe048>,\n",
       " <tensorflow.python.keras.layers.core.Dense at 0x1c60ebfef08>,\n",
       " <tensorflow.python.keras.layers.core.Dense at 0x1c60ebfeec8>,\n",
       " <tensorflow.python.keras.layers.core.Dense at 0x1c630af4848>]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "flatten (Flatten)            (None, 784)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 300)               235500    \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 100)               30100     \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                1010      \n",
      "=================================================================\n",
      "Total params: 266,610\n",
      "Trainable params: 266,610\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "keras.utils.plot_model(model, \"my_fashion_mnist_model.png\", show_shapes=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'dense'"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hidden1 = model.layers[1]\n",
    "hidden1.name"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.get_layer(hidden1.name) is hidden1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "weights, biases = hidden1.get_weights()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "       0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.], dtype=float32)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "biases"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(loss=\"sparse_categorical_crossentropy\",\n",
    "              optimizer=\"sgd\",\n",
    "              metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 55000 samples, validate on 5000 samples\n",
      "Epoch 1/30\n",
      "55000/55000 [==============================] - 3s 60us/sample - loss: 0.7226 - accuracy: 0.7641 - val_loss: 0.5073 - val_accuracy: 0.8320\n",
      "Epoch 2/30\n",
      "55000/55000 [==============================] - 3s 55us/sample - loss: 0.4843 - accuracy: 0.8322 - val_loss: 0.4543 - val_accuracy: 0.8484\n",
      "Epoch 3/30\n",
      "55000/55000 [==============================] - 3s 55us/sample - loss: 0.4414 - accuracy: 0.8462 - val_loss: 0.4379 - val_accuracy: 0.8486\n",
      "Epoch 4/30\n",
      "55000/55000 [==============================] - 3s 54us/sample - loss: 0.4129 - accuracy: 0.8547 - val_loss: 0.4156 - val_accuracy: 0.8564\n",
      "Epoch 5/30\n",
      "55000/55000 [==============================] - 3s 53us/sample - loss: 0.3926 - accuracy: 0.8619 - val_loss: 0.3811 - val_accuracy: 0.8654\n",
      "Epoch 6/30\n",
      "55000/55000 [==============================] - 3s 56us/sample - loss: 0.3771 - accuracy: 0.8668 - val_loss: 0.3727 - val_accuracy: 0.8692\n",
      "Epoch 7/30\n",
      "55000/55000 [==============================] - 3s 55us/sample - loss: 0.3624 - accuracy: 0.8731 - val_loss: 0.3713 - val_accuracy: 0.8710\n",
      "Epoch 8/30\n",
      "55000/55000 [==============================] - 3s 60us/sample - loss: 0.3516 - accuracy: 0.8750 - val_loss: 0.3664 - val_accuracy: 0.8692\n",
      "Epoch 9/30\n",
      "55000/55000 [==============================] - 3s 58us/sample - loss: 0.3421 - accuracy: 0.8775 - val_loss: 0.3440 - val_accuracy: 0.8784\n",
      "Epoch 10/30\n",
      "55000/55000 [==============================] - 3s 60us/sample - loss: 0.3326 - accuracy: 0.8811 - val_loss: 0.3497 - val_accuracy: 0.8776\n",
      "Epoch 11/30\n",
      "55000/55000 [==============================] - 3s 58us/sample - loss: 0.3241 - accuracy: 0.8837 - val_loss: 0.3357 - val_accuracy: 0.8814\n",
      "Epoch 12/30\n",
      "55000/55000 [==============================] - 3s 57us/sample - loss: 0.3158 - accuracy: 0.8867 - val_loss: 0.3314 - val_accuracy: 0.8854\n",
      "Epoch 13/30\n",
      "55000/55000 [==============================] - 3s 61us/sample - loss: 0.3072 - accuracy: 0.8900 - val_loss: 0.3322 - val_accuracy: 0.8822\n",
      "Epoch 14/30\n",
      "55000/55000 [==============================] - 3s 57us/sample - loss: 0.3014 - accuracy: 0.8921 - val_loss: 0.3234 - val_accuracy: 0.8864\n",
      "Epoch 15/30\n",
      "55000/55000 [==============================] - 3s 60us/sample - loss: 0.2952 - accuracy: 0.8937 - val_loss: 0.3170 - val_accuracy: 0.8892\n",
      "Epoch 16/30\n",
      "55000/55000 [==============================] - 3s 60us/sample - loss: 0.2896 - accuracy: 0.8967 - val_loss: 0.3241 - val_accuracy: 0.8860\n",
      "Epoch 17/30\n",
      "55000/55000 [==============================] - 3s 62us/sample - loss: 0.2832 - accuracy: 0.8985 - val_loss: 0.3175 - val_accuracy: 0.8916\n",
      "Epoch 18/30\n",
      "55000/55000 [==============================] - 3s 60us/sample - loss: 0.2780 - accuracy: 0.8998 - val_loss: 0.3101 - val_accuracy: 0.8910\n",
      "Epoch 19/30\n",
      "55000/55000 [==============================] - 3s 61us/sample - loss: 0.2729 - accuracy: 0.9020 - val_loss: 0.3179 - val_accuracy: 0.8848\n",
      "Epoch 20/30\n",
      "55000/55000 [==============================] - 3s 59us/sample - loss: 0.2679 - accuracy: 0.9045 - val_loss: 0.3225 - val_accuracy: 0.8856\n",
      "Epoch 21/30\n",
      "55000/55000 [==============================] - 3s 59us/sample - loss: 0.2634 - accuracy: 0.9046 - val_loss: 0.2984 - val_accuracy: 0.8944\n",
      "Epoch 22/30\n",
      "55000/55000 [==============================] - 3s 56us/sample - loss: 0.2575 - accuracy: 0.9072 - val_loss: 0.3093 - val_accuracy: 0.8894\n",
      "Epoch 23/30\n",
      "55000/55000 [==============================] - 3s 60us/sample - loss: 0.2538 - accuracy: 0.9089 - val_loss: 0.2991 - val_accuracy: 0.8902\n",
      "Epoch 24/30\n",
      "55000/55000 [==============================] - 3s 56us/sample - loss: 0.2491 - accuracy: 0.9105 - val_loss: 0.3096 - val_accuracy: 0.8854\n",
      "Epoch 25/30\n",
      "55000/55000 [==============================] - 4s 67us/sample - loss: 0.2451 - accuracy: 0.9127 - val_loss: 0.3092 - val_accuracy: 0.8886\n",
      "Epoch 26/30\n",
      "55000/55000 [==============================] - 3s 62us/sample - loss: 0.2407 - accuracy: 0.9142 - val_loss: 0.3251 - val_accuracy: 0.8862\n",
      "Epoch 27/30\n",
      "55000/55000 [==============================] - 3s 60us/sample - loss: 0.2366 - accuracy: 0.9150 - val_loss: 0.3108 - val_accuracy: 0.8870\n",
      "Epoch 28/30\n",
      "55000/55000 [==============================] - 3s 58us/sample - loss: 0.2322 - accuracy: 0.9175 - val_loss: 0.2922 - val_accuracy: 0.8950\n",
      "Epoch 29/30\n",
      "55000/55000 [==============================] - 3s 59us/sample - loss: 0.2286 - accuracy: 0.9185 - val_loss: 0.2971 - val_accuracy: 0.8932\n",
      "Epoch 30/30\n",
      "55000/55000 [==============================] - 3s 57us/sample - loss: 0.2256 - accuracy: 0.9189 - val_loss: 0.3032 - val_accuracy: 0.8924\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train, y_train, epochs=30,\n",
    "                    validation_data=(X_valid, y_valid))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'batch_size': 32,\n",
       " 'epochs': 30,\n",
       " 'steps': 1719,\n",
       " 'samples': 55000,\n",
       " 'verbose': 0,\n",
       " 'do_validation': True,\n",
       " 'metrics': ['loss', 'accuracy', 'val_loss', 'val_accuracy']}"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "history.params"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "history.history.keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saving figure keras_learning_curves_plot\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "pd.DataFrame(history.history).plot(figsize=(8, 5))\n",
    "plt.grid(True)\n",
    "plt.gca().set_ylim(0, 1)\n",
    "save_fig(\"keras_learning_curves_plot\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10000/10000 [==============================] - 0s 32us/sample - loss: 0.3358 - accuracy: 0.8782\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.3358254776954651, 0.8782]"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.  , 0.  , 0.  , 0.  , 0.  , 0.  , 0.  , 0.01, 0.  , 0.98],\n",
       "       [0.  , 0.  , 0.99, 0.  , 0.  , 0.  , 0.  , 0.  , 0.  , 0.  ],\n",
       "       [0.  , 1.  , 0.  , 0.  , 0.  , 0.  , 0.  , 0.  , 0.  , 0.  ]],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_new = X_test[:3]\n",
    "y_proba = model.predict(X_new)\n",
    "y_proba.round(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([9, 2, 1], dtype=int64)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred = model.predict_classes(X_new)\n",
    "y_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['Ankle boot', 'Pullover', 'Trouser'], dtype='<U11')"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.array(class_names)[y_pred]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([9, 2, 1], dtype=uint8)"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_new = y_test[:3]\n",
    "y_new"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saving figure fashion_mnist_images_plot\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAaAAAACUCAYAAADVqv1WAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8li6FKAAAX4klEQVR4nO3de5AV1Z0H8O9PAXkMD2GQx4hDEBAjClgJLj5R2CiIGtgkagxqlVIp3V13k3KLWldRt5JQYmpX3V3jmvWVTRxdFQuzkYi4vhBBd1FEeYnA8BgEeY8gCNr7x+2p3PPtw+2eywynZ/h+qqbgd/v26Z7bZ+ZMn1+fcyyKIoiIiBxpx4Q+AREROTqpARIRkSDUAImISBBqgEREJAg1QCIiEoQaIBERCaJVNUBmFpnZwMZuSynzejObd/hnJy2JmfWP60ybOH7NzG4MfV4irUkuG6D4h32HmR0X+lyai5mNNrMNoc/jaGBma83sCzP73Mw2m9ljZlYR+rwk3+L60vD1dVEd+tzMrgl9fq1B7hogM+sP4DwAEYDLg56MtCaXRVFUAeBMAN8GcHvg80llZseGPoejWRRFFQ1fANYhrkPx1+/4/Q13yyHl4RwaI3cNEIBrASwA8DiA64o3mNnjZvZvZvYHM6s3s4VmdrKvEDM718zWm9mFnm3HmdkvzWxd/BfxQ2bWocQ5mZn9i5ntMrPlZjamaENfM3vBzLab2Sozm0LHuc/M6uKv++LXOgGYDaBv0V9UfRv1KUlZoijaiMJnPzS+MxrbsM3M7jKz36aVYWbHmNntZlZrZlvM7Ddm1jXe9kcz+yt6/2IzmxT/f4iZvRzXlxVm9oOi9z1uZr8ysxfNbA+ARN2V/DCzn5nZ02ZWY2b1AH5kZu3N7AEz22RmG83sn8ysXfz+G83staL928TdvP3jeIKZLYt/t20ws58UvffyuB7tNLN5Zja0aNsGM/s7M1sCYO8R+vabRF4boN/FXxebWS/afjWAuwEcD2AVgJ9zAWZ2MYAaAH8RRdGrnmPcA2AwgOEABgKoAjCtxDmdBWA1gEoAdwKYaWbd4201ADYA6AvgewB+UdRA/QOAP4uPMwzASAC3R1G0B8A4AHVFf1HVlTi+NBEz6wdgPID3DqOY6+OvCwEMAFAB4F/jbU+iUEcbjvdNANUA/hD/4fFy/J4T4vc9aGanFZX9QxTqdGcAyj3m30QUrmdXAE+j8HvkWwDOADACwDkA/j5jWY8BuCGKos7x/q8DgJl9G8CvAdwIoAeARwHMamjYYleh8Dul62F+P0dWFEW5+QJwLoADACrjeDmAnxRtfxzAfxTF4wEsL4ojFC52LYDTqewIhcbGAOwBcHLRtlEA1hzinK4HUAfAil57B8BkAP0AfAWgc9G26QAej///CYDxRdsuBrA2/v9oABtCf+ZHwxeAtQA+B7AzrhsPAugQvz626H13Afht/P/+cZ1pE8evAbgx/v8rAG4u2u+UuN62QaHh2AOgOt72cwCPxv+/EsCbdG7/DuDOovr9m9Cfl74OWYfG0ms/A/A/9FotgO8UxZcCWBX//0YArxVtaxPXsf5xXBe/pzOV+euGOlL02icAzon/vwHAtaE/o3K+8nYHdB2AOVEUbY3jJ0HdcAA+Lfr/XhT++iz2twD+K4qiJYc4Rk8AHQH8X3w7uxPAH+PXD2VjFF/pWC0Kdzx9AWyPoqietlXF/+8bx7yfHHnfjaKoWxRF1VEU3RxF0ReHUZbvurYB0CuuC39A4S9SxP825AuqAZzVUO/iuncNgN5FZa0/jPOSI4+vVx8k60YVspmIQt57Xfwg1lnx69UAplK96UPltsh6k5uEVZyD+QGAY82soZE5DkA3MxsWRdHijEV9H8AjZrYxiqL7PNu3AvgCwGlRIR+QRZWZWVEjdBKAF1D4i6W7mXUuaoROAtBQbh0Kleejom0NXW2ahjy8PSj8MdKg96HeSBqua4OTABwEsDmOawDcaWZvoHCn1dANvB7A61EU/XmJslUvWha+XptQqBsr4rj490HJ+hZF0UIAl5tZWwB/A+ApAN9Aod7cHUXRPY04jxYhT3dA30WhO+ubKORMhgM4FcCbKOSFsqoDMAbALWZ2M2+MouhrFG5p/9nMTgAAM6uK80aHckJcXlsz+358Xi9GUbQewHwA0+Pk4xkAbsCf/uKtAXC7mfU0s0oU+ocbktybAfRoSF5LEO8DuCq+rt9CIYeXRQ2An5jZN6zwOPcvADwdRdHBePuLKPwS+sf49a/j1/8bwGAzmxwfs62ZfdvMTm26b0kCqwEwzcwqzawngDvwp5/5xQDOMLPT4z+472zYycw6mNkPzaxLFEUHANSj8PsQAB4G8JdxXTEzqzCzy+KcYouWpwboOgCPRVG0LoqiTxu+UEjuXmONeLwwiqJ1KDRCU80/eHAqCg8wLDCz3QDmotCPfygLAQxC4e7p5wC+F0XRtnjb1SjkC+oAPI9CX+3L8bafAfhfAB8AWAJgUfwaoihajkJlXR3fVqtr7si7A8DJAHag8GDLkxn3exTAfwJ4A8AaAPsA/HXDxiiK9gOYCWBscZnxXfJ3UOiWq0OhO/keFO70pXW4G4WGZgkKP/cLUcgLI4qipSj8sfIaCndIb9C+1wGojX8n3YBCnrnhzugmAL9Coa6uBPCjZv4+jghzUxsiIiJHRp7ugERE5CiiBkhERIJQAyQiIkGoARIRkSDUAImISBBpjzbrEbnWy5qx7BZRb+rr6xOvvfPOO048ZsyYxHsaa9GiRU5cUeFO3jF48ODDPsYR1OrrDT8ZbOZ+y6+88kpinwceeMCJhw8f7sSffvqpEw8cmFya7PPPP3fiHTt2OHGbNu6v6zVr1iTKeP755xOv5YS33ugOSEREglADJCIiQaQNRM3FLbE0i1bXlbJv3z4nvu8+dyrAmpoaJ+YuDgD47LPPnLhDB3eZKN8+adq3b18y5q4VADj//POdeMqUKU58ySWXNPo8mkirqzfs66+/duJjjnH/Tj/33HMT+7z11luNOkaXLl0Sr+3d6y7lc/DgQSfmuvjFF8n5dH//+9878YQJExp1Xs1IXXAiIpIfaoBERCQINUAiIhKEckBHrxbdlz916tTEaw8//LAT796924k7duzoxNynDiTzMdzPfuDAASf+6quvwI47zp3cmo/DP3P79+9PlMHH5eOMGjXKid94gydWbjYtut40hc6dOydea9u2rRP37Omub7lnzx4n9tUbzg1ymVxvVq1alSjj3nvvdeJbb7018Z5AlAMSEZH8UAMkIiJBqAESEZEg1ACJiEgQmZe5FgmJHzCYMWNG4j29e/d24k6dOjkxz+nlewCHHzJIG0TKZQLJgYs8oJBxmUByvrhjjz3WiXng42WXXZYogwclStPgOdsAoLKy0on5ARge3MoPqvjew8fx7cPWr1+f+p480R2QiIgEoQZIRESCUAMkIiJBKAckLcIdd9zhxL7JHDkfw4P9eE0Wn27dujlx2sShvnwAT4rao0ePkuflm4yUB6dyvqpXr15O7BuIunXrVifmPIVks3nz5tT38DX05QaL+fKCPPCU835cpu9nYMuWLSWPmze6AxIRkSDUAImISBBqgEREJAjlgKRF2LVrlxP7xkRwnoRzPjfddJMT//jHP06UceaZZzoxjyXasGGDE/smpqyurnZiziHwuXOZAFBVVVVyn/r6eif2LU62evVqJ1YOqDwffvhh6nvatWvnxHw9OJ/jy/vxOCCuz1nGEnHeL+90ByQiIkGoARIRkSDUAImISBDKAUmLwONifPOnpSyuiOnTpztx165dE+/hfva9e/c68ejRo5341VdfLXlMADj11FOdePny5U7M84YBwP333+/EPA6KFzzzLXA2b948Jx45cmTquUrS4sWLnZjzPUCyPnK94bFhnNMEkuPF0uYu9C1kyDnLvNMdkIiIBKEGSEREglADJCIiQagBEhGRIPQQQjPj5DAvVpY2aSGQTDbyALSPP/7YiQcNGtSYU8ylL7/8suR23+fmS8oWu/baa5141qxZqeexY8cOJ+aHDqZNm5bYhyeJfOqpp5x4+/btTlxbW5so48orr3Rifgghy4Sm77//fuI1abx3333XiflnGEg+dMDXgx864AHPQPJ6HX/88U7MP/d8TADo169f4rU80x2QiIgEoQZIRESCUAMkIiJBHLU5IB7U5RvEyH29GzdudOK3337biceNG5cooykGhvkmHSw2c+ZMJ546dephHzO0urq6ktt9/fC+CTmL+Sb9TPPMM8+U3D558uTEax06dHBiztcMGzbMiTdt2pQoo6KiIuspHhLnBqU8y5Ytc2JeOA5I1kdeqLBPnz5OvGDBgkQZnNfkQdEc+xa16969e+K1PNMdkIiIBKEGSEREglADJCIiQRy1OSDmyymwN99804kXLlzoxL68xS233HJ4JwZgy5YtTvzSSy85sW9RtJbus88+a/Q+3CfOffV8fbhP3eeCCy4ouf3iiy9OvLZmzRon5n752bNnOzFPcAok80ScE+Jz5wXPgOSCfFIeHsPj+6zTckCTJk1q9HG5Pnfs2DF1n7Txc3mjOyAREQlCDZCIiAShBkhERII4anNAWebS4jmgeDxAr169nNg37mLixIlOzPM78UJV1dXViTK2bdvmxLyAWVVVVWKflo7HXLG0xeeAZJ8550R8eT8ud8WKFU7MY6xWr16deh5pC9KtW7cusc+DDz7oxDxuJG2eMCD9M5RsNm/e7MTljO27+uqrU9/D15DnDKysrEwtwzc/XJ7pDkhERIJQAyQiIkGoARIRkSDUAImISBBHzUMIPHCPHzrYs2dPYp9nn33WiTlJyA8Q1NfXJ8pIm/SU448++ihRxoknnujEnIDmBypag7SBqL7BgDxwj2MezHnbbbelljFnzhwnXrx4sRP7rhc/JMIPHfCDDLz4HJC+mBzXZ98CfQcOHChZhmTDk9z6Bn6n/QxeeOGFqccZNWqUE/Nkx77JR1mPHj1S35MnugMSEZEg1ACJiEgQaoBERCSI4Dkg34DCtIWZeLuv/5v7ZH05g2IPPfRQ4jUeaNq+fXsnrq2tdWLOCfnK4H5cPnffIDfOPfHkiPv373diXz6rKRbGO5J8i7QVyzKIlD/rrl27OvH06dNTz4P34eu5dOnS1DJ69+7txFu3bnVirldZZBlInbZP2s+EZMf5Nr4eaYtKAkD//v2deN68eU6cZfA119e80x2QiIgEoQZIRESCUAMkIiJBNHsOiPsts+RvWNpicb5n8NP6t2tqapzYt3jXiBEjnJhzCjt37nRiXngMSD6Xz/3/vHBVlmf9+TPlCQh9k6IOHz48tdw8KWdBunbt2jnxRRdd5MS8oCCPrwKS9Ybza1zXeGyRD19TziPxMXzlduvWzYl5nJCv7rG1a9c68cknn5y6jyT5fmfxQnDlfLZcH7muZfld2dLoDkhERIJQAyQiIkGoARIRkSCaPQeU1m/JY3x8r3G/PJeZZTzDo48+6sQrV6504n79+iX24YXgOPfCc0T5Fobj+eH43HnRNN9YorQ8GnvppZcSr7W0HBDn15hv3j3+/K+//nonnj17thPzZ+/DddFXX9Pw9eKckC8HxONIJk2a5MRpc8X5cP5ROaDy+MZc8di70047rdHljh8/3olnzJjhxOXUvbzTHZCIiAShBkhERIJQAyQiIkGoARIRkSAO6yGELEkxTsByQt03yDRt4Cmrq6tLvDZz5kwn5gcGBg0a5MQ8IBRIJof5oYS2bds6se/hAB4kyvh79U1ayO/hiUX5uG+99VbJY7YE/Fkzvp4AcMIJJzgxL9zH+PoB6ZPFNrZu+srIMsCQ695ZZ51V8hi+8+JJTltjEjsE38B3/r02YMCARpc7bNgwJ+bBrVkGqbe0SYd1ByQiIkGoARIRkSDUAImISBAlc0BpC1g1RX+4D09EyZMorlixwol9i5fxxJRdunRxYh7ouHv37kQZvMgU98vz58HnCST7bXlSST7PLP3LHTp0KLmPb4LMDz/80ImHDh2aeE+e8PXhfIZvwC73fy9btqzkMXwDCvmas3ImhCxnQl7+/ssZ0M3H5YGokg1PEupb8JF/F/bt27fRx0lbVFA5IBERkSaiBkhERIJQAyQiIkGU7HRMm+Rz8+bNiddqa2udmPtLOfaN51izZo0T81ga7ivt3LlzogzuE9+1a1fJ4/r6X/m4nHvhMTv83D4A9OnTx4k518TH8I1d4TFK27dvd2LO+fgW1+N98q6cMSunnHKKE3/yyScl3+/Lq/Bx08axZZE2Galv7Bcfh8c4sSw5oHIW+ZPkZ7969erEe/ia8mTHWXA+mKXliID0cYd5ozsgEREJQg2QiIgEoQZIRESCaNRccHPnznVi3xxs3E/J/c5pY4t8ZXCOh3MivpwH93/zGB7Otfj60Pk4fO78zL1v/A2P+ymnH57PlccccD7Ll4vK0n+cJzweJ8v5cw7o9ddfL/n+LOMquB5xPckyFo7L4DjLgoo8FoXjLGN8fPMdSrqRI0c6sW98GefxylkwMI1v4cK088g73QGJiEgQaoBERCQINUAiIhKEGiAREQmiZGZ3zpw5TvzII4848ZAhQxL78MBLfoCAk7i+wVec7OekLZfpS7pzcri+vr5kmb4BsWkLifHDD76BuUuXLi15rr7JRxk/3MCDeXmiTt/DEGkDGfOGB/1mSdTzNV++fLkT8wJ0WT77cqQtOMdxlgcsVq1a5cS9e/d2Yt+DOPz9trRBinlx/vnnO/Fjjz2WeA//HnvvvfcO+7hcn7M8NFPOBNEhtayzFRGRVkMNkIiIBKEGSEREgijZ+cwDsBYsWODES5YsSewzb968kgfkfmnfRKLdu3cvGXft2tWJfTkgzvFs27bNiXlRO1//OE8cyn33ixcvduIzzjgjUUb//v2d+OWXX3ZiHlyWpQ+Xcwa8+BUvvgckc2B5x99jlnwND17lCVg7duzoxOVMeMrKWaCO81lZ+vZnzZrlxFyvFi1alNiH69KOHTsynqEUO/vss52Yc65A8po2Rc6Vf46zTITbFHX6SNIdkIiIBKEGSEREglADJCIiQZTMAfFEmtOmTUstkCc8XLhwoRNz7mX+/PmJMtauXevEH3zwgRPzOBhf3yj3zXN/OOeVTj/99EQZY8eOdeLx48c7sa8vOM3ll1/uxOvWrXPiHj16JPbhvmDOm3G+xDch4eDBgxt1nqHx9dq3b1/qPjzuh/Nr/LlwzghI9uWn9bv7tvNraXmiLP32/DPB+cZnn302sQ8f1/f9Srrq6mon9uVYua5xfeVF7AYMGJB6XM6XZ7l+zTW2rbnoDkhERIJQAyQiIkGoARIRkSCafJUynodszJgxJeObb765qU8h11544YXQp9AicL4mS56Ex7lwPzyXWc78chz78jtpc7+lLVAHJMe6vf32206cJafHx/XNdyiN51sYjsdy8djEcnJAPK8m5wF5oUpAOSAREZFM1ACJiEgQaoBERCQINUAiIhJEkz+EINIUeBAeTyTKA54B4Kc//akTz50714k5CV/O4l1pDxgA6YNX+YEK33ns2rXLiUePHu3EEyZMcOK77747UQY/ZOFLnktS2kDiiRMnJvZ58sknnZivMU/SzIPcfbjOp50n4H8wIc90ByQiIkGoARIRkSDUAImISBDKAUku8YSznM/gHBGQnKyxZ8+eTvzxxx87sW8wYHMs6JWWU/B9Lzyolhc4q6ysTD0u55Zqa2tT95H063XFFVck9nniiSecuF27dk783HPPOfFdd92Veh48qDRL/tE3EXGe6Q5IRESCUAMkIiJBqAESEZEglAOSXDrnnHOcmCfj9C0GyBN0rly5sulPLCd4cktepBBIjvsZOXJks55Ta5E2TmvcuHGJfXj8DX/25Yw5Gzp0qBMvWbLEiX0/A5s2bWr0cULSHZCIiAShBkhERIJQAyQiIkEoByS5xPkKnseNx1kA5fWzt1Q85sk3zxsvitapU6dmPafWIstChay6utqJFyxY4MR79+514vnz5yfKOPvss52YxwHxAot8fQFg69at6SebI0fPT6yIiOSKGiAREQlCDZCIiAShBkhERILQQwiSS1VVVU48YsQIJ/YNwktLsh88eNCJfcnmtMXkjhQ+Dz7XgQMHOvGll16aKGPnzp1OPGrUqCY6u9bNN8lnmilTpjjxkCFDnPiqq65yYn7gwGfy5MlOzIsUVlRUJPY577zzUsvNE90BiYhIEGqAREQkCDVAIiIShOWlz1tERI4uugMSEZEg1ACJiEgQaoBERCQINUAiIhKEGiAREQlCDZCIiATx/+TVl8RORHG7AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 518.4x172.8 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(7.2, 2.4))\n",
    "for index, image in enumerate(X_new):\n",
    "    plt.subplot(1, 3, index + 1)\n",
    "    plt.imshow(image, cmap=\"binary\", interpolation=\"nearest\")\n",
    "    plt.axis('off')\n",
    "    plt.title(class_names[y_test[index]], fontsize=12)\n",
    "plt.subplots_adjust(wspace=0.2, hspace=0.5)\n",
    "save_fig('fashion_mnist_images_plot', tight_layout=False)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Regression MLP"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.datasets import fetch_california_housing\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "housing = fetch_california_housing()\n",
    "\n",
    "X_train_full, X_test, y_train_full, y_test = train_test_split(housing.data, housing.target, random_state=42)\n",
    "X_train, X_valid, y_train, y_valid = train_test_split(X_train_full, y_train_full, random_state=42)\n",
    "\n",
    "scaler = StandardScaler()\n",
    "X_train = scaler.fit_transform(X_train)\n",
    "X_valid = scaler.transform(X_valid)\n",
    "X_test = scaler.transform(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(42)\n",
    "tf.random.set_seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 11610 samples, validate on 3870 samples\n",
      "Epoch 1/20\n",
      "11610/11610 [==============================] - 1s 52us/sample - loss: 1.6205 - val_loss: 2.0374\n",
      "Epoch 2/20\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.7162 - val_loss: 0.6571\n",
      "Epoch 3/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.6356 - val_loss: 0.5996\n",
      "Epoch 4/20\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.5989 - val_loss: 0.5662\n",
      "Epoch 5/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.5713 - val_loss: 0.5489\n",
      "Epoch 6/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.5491 - val_loss: 0.5204\n",
      "Epoch 7/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.5301 - val_loss: 0.5018\n",
      "Epoch 8/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.5142 - val_loss: 0.4815\n",
      "Epoch 9/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.5004 - val_loss: 0.4695\n",
      "Epoch 10/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.4883 - val_loss: 0.4605\n",
      "Epoch 11/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.4786 - val_loss: 0.4495\n",
      "Epoch 12/20\n",
      "11610/11610 [==============================] - 0s 35us/sample - loss: 0.4697 - val_loss: 0.4382\n",
      "Epoch 13/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.4621 - val_loss: 0.4309\n",
      "Epoch 14/20\n",
      "11610/11610 [==============================] - 0s 35us/sample - loss: 0.4556 - val_loss: 0.4247\n",
      "Epoch 15/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.4497 - val_loss: 0.4200\n",
      "Epoch 16/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.4443 - val_loss: 0.4149\n",
      "Epoch 17/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.4397 - val_loss: 0.4108\n",
      "Epoch 18/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.4354 - val_loss: 0.4059\n",
      "Epoch 19/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.4315 - val_loss: 0.4003\n",
      "Epoch 20/20\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.4281 - val_loss: 0.3981\n",
      "5160/5160 [==============================] - 0s 20us/sample - loss: 0.4218\n"
     ]
    }
   ],
   "source": [
    "model = keras.models.Sequential([\n",
    "    keras.layers.Dense(30, activation=\"relu\", input_shape=X_train.shape[1:]),\n",
    "    keras.layers.Dense(1)\n",
    "])\n",
    "model.compile(loss=\"mean_squared_error\", optimizer=keras.optimizers.SGD(lr=1e-3))\n",
    "history = model.fit(X_train, y_train, epochs=20, validation_data=(X_valid, y_valid))\n",
    "mse_test = model.evaluate(X_test, y_test)\n",
    "X_new = X_test[:3]\n",
    "y_pred = model.predict(X_new)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(pd.DataFrame(history.history))\n",
    "plt.grid(True)\n",
    "plt.gca().set_ylim(0, 1)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.37310064],\n",
       "       [1.6790789 ],\n",
       "       [3.0817137 ]], dtype=float32)"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Functional API"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(42)\n",
    "tf.random.set_seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "input_ = keras.layers.Input(shape=X_train.shape[1:])\n",
    "hidden1 = keras.layers.Dense(30, activation=\"relu\")(input_)\n",
    "hidden2 = keras.layers.Dense(30, activation=\"relu\")(hidden1)\n",
    "concat = keras.layers.concatenate([input_, hidden2])\n",
    "output = keras.layers.Dense(1)(concat)\n",
    "model = keras.models.Model(inputs=[input_], outputs=[output])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_1 (InputLayer)            [(None, 8)]          0                                            \n",
      "__________________________________________________________________________________________________\n",
      "dense_5 (Dense)                 (None, 30)           270         input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_6 (Dense)                 (None, 30)           930         dense_5[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "concatenate (Concatenate)       (None, 38)           0           input_1[0][0]                    \n",
      "                                                                 dense_6[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_7 (Dense)                 (None, 1)            39          concatenate[0][0]                \n",
      "==================================================================================================\n",
      "Total params: 1,239\n",
      "Trainable params: 1,239\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "keras.utils.plot_model(model, \"1.png\", show_shapes=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 11610 samples, validate on 3870 samples\n",
      "Epoch 1/20\n",
      "11610/11610 [==============================] - 1s 56us/sample - loss: 1.2390 - val_loss: 0.6566\n",
      "Epoch 2/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.6312 - val_loss: 0.6734\n",
      "Epoch 3/20\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.5886 - val_loss: 0.5574\n",
      "Epoch 4/20\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.5595 - val_loss: 0.5235\n",
      "Epoch 5/20\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.5361 - val_loss: 0.5011\n",
      "Epoch 6/20\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.5178 - val_loss: 0.5065\n",
      "Epoch 7/20\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.5016 - val_loss: 0.4699\n",
      "Epoch 8/20\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.4888 - val_loss: 0.4745\n",
      "Epoch 9/20\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.4772 - val_loss: 0.4425\n",
      "Epoch 10/20\n",
      "11610/11610 [==============================] - 0s 43us/sample - loss: 0.4673 - val_loss: 0.4384\n",
      "Epoch 11/20\n",
      "11610/11610 [==============================] - 0s 41us/sample - loss: 0.4586 - val_loss: 0.4533\n",
      "Epoch 12/20\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.4504 - val_loss: 0.4179\n",
      "Epoch 13/20\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.4435 - val_loss: 0.4137\n",
      "Epoch 14/20\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.4376 - val_loss: 0.4062\n",
      "Epoch 15/20\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.4318 - val_loss: 0.4541\n",
      "Epoch 16/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4266 - val_loss: 0.3952\n",
      "Epoch 17/20\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.4221 - val_loss: 0.3910\n",
      "Epoch 18/20\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.4173 - val_loss: 0.4205\n",
      "Epoch 19/20\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.4132 - val_loss: 0.3830\n",
      "Epoch 20/20\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.4096 - val_loss: 0.3923\n",
      "5160/5160 [==============================] - 0s 21us/sample - loss: 0.4042\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss=\"mean_squared_error\", optimizer=keras.optimizers.SGD(lr=1e-3))\n",
    "history = model.fit(X_train, y_train, epochs=20,\n",
    "                    validation_data=(X_valid, y_valid))\n",
    "mse_test = model.evaluate(X_test, y_test)\n",
    "y_pred = model.predict(X_new)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Another example: we used features 0 to 4 for the wide path, features 2 to 7 for the deep path."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(42)\n",
    "tf.random.set_seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "input_A = keras.layers.Input(shape=[5], name=\"wide_input\")\n",
    "input_B = keras.layers.Input(shape=[6], name=\"deep_input\")\n",
    "hidden1 = keras.layers.Dense(30, activation=\"relu\")(input_B)\n",
    "hidden2 = keras.layers.Dense(30, activation=\"relu\")(hidden1)\n",
    "concat = keras.layers.concatenate([input_A, hidden2])\n",
    "output = keras.layers.Dense(1, name=\"output\")(concat)\n",
    "model = keras.models.Model(inputs=[input_A, input_B], outputs=[output])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 11610 samples, validate on 3870 samples\n",
      "Epoch 1/20\n",
      "11610/11610 [==============================] - 1s 57us/sample - loss: 1.8127 - val_loss: 2.1165\n",
      "Epoch 2/20\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.6852 - val_loss: 0.6178\n",
      "Epoch 3/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.5965 - val_loss: 0.5600\n",
      "Epoch 4/20\n",
      "11610/11610 [==============================] - 0s 41us/sample - loss: 0.5587 - val_loss: 0.5269\n",
      "Epoch 5/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.5321 - val_loss: 0.5185\n",
      "Epoch 6/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.5129 - val_loss: 0.4803\n",
      "Epoch 7/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4959 - val_loss: 0.4689\n",
      "Epoch 8/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4837 - val_loss: 0.4498\n",
      "Epoch 9/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4734 - val_loss: 0.4387\n",
      "Epoch 10/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4646 - val_loss: 0.4306\n",
      "Epoch 11/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4571 - val_loss: 0.4262\n",
      "Epoch 12/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4507 - val_loss: 0.4173\n",
      "Epoch 13/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4456 - val_loss: 0.4124\n",
      "Epoch 14/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4422 - val_loss: 0.4084\n",
      "Epoch 15/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4386 - val_loss: 0.4351\n",
      "Epoch 16/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4361 - val_loss: 0.4017\n",
      "Epoch 17/20\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.4326 - val_loss: 0.3990\n",
      "Epoch 18/20\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4296 - val_loss: 0.4148\n",
      "Epoch 19/20\n",
      "11610/11610 [==============================] - 0s 41us/sample - loss: 0.4278 - val_loss: 0.3957\n",
      "Epoch 20/20\n",
      "11610/11610 [==============================] - 0s 41us/sample - loss: 0.4259 - val_loss: 0.3976\n",
      "5160/5160 [==============================] - 0s 22us/sample - loss: 0.4202\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss=\"mse\", optimizer=keras.optimizers.SGD(lr=1e-3))\n",
    "\n",
    "X_train_A, X_train_B = X_train[:, :5], X_train[:, 2:]\n",
    "X_valid_A, X_valid_B = X_valid[:, :5], X_valid[:, 2:]\n",
    "X_test_A, X_test_B = X_test[:, :5], X_test[:, 2:]\n",
    "X_new_A, X_new_B = X_test_A[:3], X_test_B[:3]\n",
    "\n",
    "history = model.fit((X_train_A, X_train_B), y_train, epochs=20,\n",
    "                    validation_data=((X_valid_A, X_valid_B), y_valid))\n",
    "mse_test = model.evaluate((X_test_A, X_test_B), y_test)\n",
    "y_pred = model.predict((X_new_A, X_new_B))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Adding an auxiliary output for regularization:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "input_A = keras.layers.Input(shape=[5], name=\"wide_input\")\n",
    "input_B = keras.layers.Input(shape=[6], name=\"deep_input\")\n",
    "hidden1 = keras.layers.Dense(30, activation=\"relu\")(input_B)\n",
    "hidden2 = keras.layers.Dense(30, activation=\"relu\")(hidden1)\n",
    "concat = keras.layers.concatenate([input_A, hidden2])\n",
    "output = keras.layers.Dense(1, name=\"main_output\")(concat)\n",
    "aux_output = keras.layers.Dense(1, name=\"aux_output\")(hidden2)\n",
    "model = keras.models.Model(inputs=[input_A, input_B],\n",
    "                           outputs=[output, aux_output])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(loss=[\"mse\", \"mse\"], loss_weights=[0.9, 0.1], optimizer=keras.optimizers.SGD(lr=1e-3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 11610 samples, validate on 3870 samples\n",
      "Epoch 1/20\n",
      "11610/11610 [==============================] - 1s 74us/sample - loss: 2.0867 - main_output_loss: 1.8375 - aux_output_loss: 4.3229 - val_loss: 1.7305 - val_main_output_loss: 1.1226 - val_aux_output_loss: 7.1969\n",
      "Epoch 2/20\n",
      "11610/11610 [==============================] - 1s 49us/sample - loss: 0.9835 - main_output_loss: 0.7834 - aux_output_loss: 2.7860 - val_loss: 1.3301 - val_main_output_loss: 0.6998 - val_aux_output_loss: 6.9993\n",
      "Epoch 3/20\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.8089 - main_output_loss: 0.6704 - aux_output_loss: 2.0544 - val_loss: 1.2302 - val_main_output_loss: 0.6118 - val_aux_output_loss: 6.7919\n",
      "Epoch 4/20\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.7213 - main_output_loss: 0.6127 - aux_output_loss: 1.6975 - val_loss: 1.1911 - val_main_output_loss: 0.5980 - val_aux_output_loss: 6.5246\n",
      "Epoch 5/20\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.6698 - main_output_loss: 0.5756 - aux_output_loss: 1.5177 - val_loss: 1.1624 - val_main_output_loss: 0.6085 - val_aux_output_loss: 6.1427\n",
      "Epoch 6/20\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.6362 - main_output_loss: 0.5488 - aux_output_loss: 1.4218 - val_loss: 1.0985 - val_main_output_loss: 0.5917 - val_aux_output_loss: 5.6563\n",
      "Epoch 7/20\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.6098 - main_output_loss: 0.5269 - aux_output_loss: 1.3567 - val_loss: 1.0232 - val_main_output_loss: 0.5623 - val_aux_output_loss: 5.1668\n",
      "Epoch 8/20\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.5894 - main_output_loss: 0.5096 - aux_output_loss: 1.3054 - val_loss: 0.9374 - val_main_output_loss: 0.5192 - val_aux_output_loss: 4.6985\n",
      "Epoch 9/20\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.5725 - main_output_loss: 0.4953 - aux_output_loss: 1.2661 - val_loss: 0.8718 - val_main_output_loss: 0.4953 - val_aux_output_loss: 4.2574\n",
      "Epoch 10/20\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.5569 - main_output_loss: 0.4819 - aux_output_loss: 1.2310 - val_loss: 0.8132 - val_main_output_loss: 0.4748 - val_aux_output_loss: 3.8561\n",
      "Epoch 11/20\n",
      "11610/11610 [==============================] - 1s 43us/sample - loss: 0.5440 - main_output_loss: 0.4715 - aux_output_loss: 1.1962 - val_loss: 0.7526 - val_main_output_loss: 0.4471 - val_aux_output_loss: 3.4989\n",
      "Epoch 12/20\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.5338 - main_output_loss: 0.4639 - aux_output_loss: 1.1641 - val_loss: 0.7097 - val_main_output_loss: 0.4333 - val_aux_output_loss: 3.1952\n",
      "Epoch 13/20\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.5257 - main_output_loss: 0.4574 - aux_output_loss: 1.1390 - val_loss: 0.6748 - val_main_output_loss: 0.4271 - val_aux_output_loss: 2.9013\n",
      "Epoch 14/20\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.5194 - main_output_loss: 0.4534 - aux_output_loss: 1.1155 - val_loss: 0.6445 - val_main_output_loss: 0.4195 - val_aux_output_loss: 2.6675\n",
      "Epoch 15/20\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.5139 - main_output_loss: 0.4494 - aux_output_loss: 1.0929 - val_loss: 0.6224 - val_main_output_loss: 0.4135 - val_aux_output_loss: 2.5006\n",
      "Epoch 16/20\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.5089 - main_output_loss: 0.4462 - aux_output_loss: 1.0737 - val_loss: 0.6065 - val_main_output_loss: 0.4097 - val_aux_output_loss: 2.3759\n",
      "Epoch 17/20\n",
      "11610/11610 [==============================] - 1s 46us/sample - loss: 0.5040 - main_output_loss: 0.4429 - aux_output_loss: 1.0534 - val_loss: 0.5906 - val_main_output_loss: 0.4062 - val_aux_output_loss: 2.2484\n",
      "Epoch 18/20\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.4992 - main_output_loss: 0.4395 - aux_output_loss: 1.0366 - val_loss: 0.5802 - val_main_output_loss: 0.4039 - val_aux_output_loss: 2.1660\n",
      "Epoch 19/20\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.4956 - main_output_loss: 0.4375 - aux_output_loss: 1.0185 - val_loss: 0.5726 - val_main_output_loss: 0.4046 - val_aux_output_loss: 2.0823\n",
      "Epoch 20/20\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.4922 - main_output_loss: 0.4355 - aux_output_loss: 1.0024 - val_loss: 0.5583 - val_main_output_loss: 0.4000 - val_aux_output_loss: 1.9814\n"
     ]
    }
   ],
   "source": [
    "history = model.fit([X_train_A, X_train_B], [y_train, y_train], epochs=20,\n",
    "                    validation_data=([X_valid_A, X_valid_B], [y_valid, y_valid]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5160/5160 [==============================] - 0s 25us/sample - loss: 0.4822 - main_output_loss: 0.4277 - aux_output_loss: 0.9766\n"
     ]
    }
   ],
   "source": [
    "total_loss, main_loss, aux_loss = model.evaluate(\n",
    "    [X_test_A, X_test_B], [y_test, y_test])\n",
    "y_pred_main, y_pred_aux = model.predict([X_new_A, X_new_B])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# The subclassing API"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "class WideAndDeepModel(keras.models.Model):\n",
    "    def __init__(self, units=30, activation=\"relu\", **kwargs):\n",
    "        super().__init__(**kwargs)\n",
    "        self.hidden1 = keras.layers.Dense(units, activation=activation)\n",
    "        self.hidden2 = keras.layers.Dense(units, activation=activation)\n",
    "        self.main_output = keras.layers.Dense(1)\n",
    "        self.aux_output = keras.layers.Dense(1)\n",
    "        \n",
    "    def call(self, inputs):\n",
    "        input_A, input_B = inputs\n",
    "        hidden1 = self.hidden1(input_B)\n",
    "        hidden2 = self.hidden2(hidden1)\n",
    "        concat = keras.layers.concatenate([input_A, hidden2])\n",
    "        main_output = self.main_output(concat)\n",
    "        aux_output = self.aux_output(hidden2)\n",
    "        return main_output, aux_output\n",
    "\n",
    "model = WideAndDeepModel(30, activation=\"relu\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:AutoGraph could not transform <bound method WideAndDeepModel.call of <__main__.WideAndDeepModel object at 0x000001C6147C7588>> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: Bad argument number for Name: 3, expecting 4\n",
      "WARNING: AutoGraph could not transform <bound method WideAndDeepModel.call of <__main__.WideAndDeepModel object at 0x000001C6147C7588>> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: Bad argument number for Name: 3, expecting 4\n",
      "Train on 11610 samples, validate on 3870 samples\n",
      "Epoch 1/10\n",
      "11610/11610 [==============================] - 1s 87us/sample - loss: 2.4897 - output_1_loss: 2.3161 - output_2_loss: 4.0451 - val_loss: 2.0961 - val_output_1_loss: 1.8751 - val_output_2_loss: 4.0797\n",
      "Epoch 2/10\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.9934 - output_1_loss: 0.8363 - output_2_loss: 2.4100 - val_loss: 0.9892 - val_output_1_loss: 0.7246 - val_output_2_loss: 3.3681\n",
      "Epoch 3/10\n",
      "11610/11610 [==============================] - 0s 43us/sample - loss: 0.7877 - output_1_loss: 0.6759 - output_2_loss: 1.7935 - val_loss: 0.8825 - val_output_1_loss: 0.6237 - val_output_2_loss: 3.2092\n",
      "Epoch 4/10\n",
      "11610/11610 [==============================] - 0s 43us/sample - loss: 0.6950 - output_1_loss: 0.6060 - output_2_loss: 1.4945 - val_loss: 0.7994 - val_output_1_loss: 0.5576 - val_output_2_loss: 2.9738\n",
      "Epoch 5/10\n",
      "11610/11610 [==============================] - 0s 43us/sample - loss: 0.6378 - output_1_loss: 0.5604 - output_2_loss: 1.3345 - val_loss: 0.7344 - val_output_1_loss: 0.5288 - val_output_2_loss: 2.5834\n",
      "Epoch 6/10\n",
      "11610/11610 [==============================] - 0s 43us/sample - loss: 0.5984 - output_1_loss: 0.5270 - output_2_loss: 1.2401 - val_loss: 0.6732 - val_output_1_loss: 0.4856 - val_output_2_loss: 2.3606\n",
      "Epoch 7/10\n",
      "11610/11610 [==============================] - 0s 43us/sample - loss: 0.5683 - output_1_loss: 0.5006 - output_2_loss: 1.1785 - val_loss: 0.6313 - val_output_1_loss: 0.4691 - val_output_2_loss: 2.0896\n",
      "Epoch 8/10\n",
      "11610/11610 [==============================] - 1s 43us/sample - loss: 0.5451 - output_1_loss: 0.4802 - output_2_loss: 1.1281 - val_loss: 0.5931 - val_output_1_loss: 0.4519 - val_output_2_loss: 1.8622\n",
      "Epoch 9/10\n",
      "11610/11610 [==============================] - 0s 43us/sample - loss: 0.5282 - output_1_loss: 0.4658 - output_2_loss: 1.0888 - val_loss: 0.5780 - val_output_1_loss: 0.4525 - val_output_2_loss: 1.7062\n",
      "Epoch 10/10\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.5159 - output_1_loss: 0.4558 - output_2_loss: 1.0555 - val_loss: 0.5602 - val_output_1_loss: 0.4489 - val_output_2_loss: 1.5605\n",
      "5160/5160 [==============================] - 0s 24us/sample - loss: 0.5009 - output_1_loss: 0.4438 - output_2_loss: 1.0233\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss=\"mse\", loss_weights=[0.9, 0.1], optimizer=keras.optimizers.SGD(lr=1e-3))\n",
    "history = model.fit((X_train_A, X_train_B), (y_train, y_train), epochs=10,\n",
    "                    validation_data=((X_valid_A, X_valid_B), (y_valid, y_valid)))\n",
    "total_loss, main_loss, aux_loss = model.evaluate((X_test_A, X_test_B), (y_test, y_test))\n",
    "y_pred_main, y_pred_aux = model.predict((X_new_A, X_new_B))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Saving and Restoring"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(42)\n",
    "tf.random.set_seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.models.Sequential([\n",
    "    keras.layers.Dense(30, activation=\"relu\", input_shape=[8]),\n",
    "    keras.layers.Dense(30, activation=\"relu\"),\n",
    "    keras.layers.Dense(1)\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 11610 samples, validate on 3870 samples\n",
      "Epoch 1/10\n",
      "11610/11610 [==============================] - 1s 57us/sample - loss: 1.8423 - val_loss: 5.2165\n",
      "Epoch 2/10\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.6876 - val_loss: 0.7732\n",
      "Epoch 3/10\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.5954 - val_loss: 0.5446\n",
      "Epoch 4/10\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.5553 - val_loss: 0.5425\n",
      "Epoch 5/10\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.5268 - val_loss: 0.5539\n",
      "Epoch 6/10\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.5049 - val_loss: 0.4701\n",
      "Epoch 7/10\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.4852 - val_loss: 0.4562\n",
      "Epoch 8/10\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.4706 - val_loss: 0.4452\n",
      "Epoch 9/10\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.4576 - val_loss: 0.4406\n",
      "Epoch 10/10\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.4476 - val_loss: 0.4185\n",
      "5160/5160 [==============================] - 0s 20us/sample - loss: 0.4376\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss=\"mse\", optimizer=keras.optimizers.SGD(lr=1e-3))\n",
    "history = model.fit(X_train, y_train, epochs=10, validation_data=(X_valid, y_valid))\n",
    "mse_test = model.evaluate(X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save(\"my_keras_model.h5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.models.load_model(\"my_keras_model.h5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.551559 ],\n",
       "       [1.6555369],\n",
       "       [3.0014234]], dtype=float32)"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.predict(X_new)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save_weights(\"my_keras_weights_1.ckpt\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.training.tracking.util.CheckpointLoadStatus at 0x1c6100ec948>"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.load_weights(\"my_keras_weights_1.ckpt\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Using Callbacks during Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "keras.backend.clear_session()\n",
    "np.random.seed(42)\n",
    "tf.random.set_seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.models.Sequential([\n",
    "    keras.layers.Dense(30, activation=\"relu\", input_shape=[8]),\n",
    "    keras.layers.Dense(30, activation=\"relu\"),\n",
    "    keras.layers.Dense(1)\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 11610 samples, validate on 3870 samples\n",
      "Epoch 1/10\n",
      "11610/11610 [==============================] - 1s 56us/sample - loss: 1.8423 - val_loss: 5.2165\n",
      "Epoch 2/10\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.6876 - val_loss: 0.7732\n",
      "Epoch 3/10\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.5954 - val_loss: 0.5446\n",
      "Epoch 4/10\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.5553 - val_loss: 0.5425\n",
      "Epoch 5/10\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.5268 - val_loss: 0.5539\n",
      "Epoch 6/10\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.5049 - val_loss: 0.4701\n",
      "Epoch 7/10\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4852 - val_loss: 0.4562\n",
      "Epoch 8/10\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4706 - val_loss: 0.4452\n",
      "Epoch 9/10\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.4576 - val_loss: 0.4406\n",
      "Epoch 10/10\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4476 - val_loss: 0.4185\n",
      "5160/5160 [==============================] - 0s 27us/sample - loss: 0.4376\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss=\"mse\", optimizer=keras.optimizers.SGD(lr=1e-3))\n",
    "checkpoint_cb = keras.callbacks.ModelCheckpoint(\"my_keras_model.h5\", save_best_only=True)\n",
    "history = model.fit(X_train, y_train, epochs=10,\n",
    "                    validation_data=(X_valid, y_valid),\n",
    "                    callbacks=[checkpoint_cb])\n",
    "model = keras.models.load_model(\"my_keras_model.h5\") # rollback to best model\n",
    "mse_test = model.evaluate(X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 11610 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "11610/11610 [==============================] - 1s 54us/sample - loss: 0.4385 - val_loss: 0.4287\n",
      "Epoch 2/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.4319 - val_loss: 0.4117\n",
      "Epoch 3/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.4252 - val_loss: 0.3975\n",
      "Epoch 4/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.4199 - val_loss: 0.3943\n",
      "Epoch 5/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.4150 - val_loss: 0.3964\n",
      "Epoch 6/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.4110 - val_loss: 0.3907\n",
      "Epoch 7/100\n",
      "11610/11610 [==============================] - 0s 41us/sample - loss: 0.4070 - val_loss: 0.3823\n",
      "Epoch 8/100\n",
      "11610/11610 [==============================] - 0s 43us/sample - loss: 0.4036 - val_loss: 0.3786\n",
      "Epoch 9/100\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4003 - val_loss: 0.3739\n",
      "Epoch 10/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3973 - val_loss: 0.3724\n",
      "Epoch 11/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3944 - val_loss: 0.3697\n",
      "Epoch 12/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3916 - val_loss: 0.3670\n",
      "Epoch 13/100\n",
      "11610/11610 [==============================] - 0s 41us/sample - loss: 0.3892 - val_loss: 0.3638\n",
      "Epoch 14/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3869 - val_loss: 0.3633\n",
      "Epoch 15/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3845 - val_loss: 0.4051\n",
      "Epoch 16/100\n",
      "11610/11610 [==============================] - 0s 41us/sample - loss: 0.3826 - val_loss: 0.3662\n",
      "Epoch 17/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3805 - val_loss: 0.3554\n",
      "Epoch 18/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3782 - val_loss: 0.3684\n",
      "Epoch 19/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3762 - val_loss: 0.3523\n",
      "Epoch 20/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3744 - val_loss: 0.3622\n",
      "Epoch 21/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3727 - val_loss: 0.3498\n",
      "Epoch 22/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3709 - val_loss: 0.3651\n",
      "Epoch 23/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3694 - val_loss: 0.3538\n",
      "Epoch 24/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3681 - val_loss: 0.3479\n",
      "Epoch 25/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3668 - val_loss: 0.3475\n",
      "Epoch 26/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3654 - val_loss: 0.3483\n",
      "Epoch 27/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3641 - val_loss: 0.3787\n",
      "Epoch 28/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3631 - val_loss: 0.3461\n",
      "Epoch 29/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3618 - val_loss: 0.3709\n",
      "Epoch 30/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3607 - val_loss: 0.3530\n",
      "Epoch 31/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3596 - val_loss: 0.3429\n",
      "Epoch 32/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3586 - val_loss: 0.3441\n",
      "Epoch 33/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3577 - val_loss: 0.3534\n",
      "Epoch 34/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3567 - val_loss: 0.3359\n",
      "Epoch 35/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3557 - val_loss: 0.3828\n",
      "Epoch 36/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3551 - val_loss: 0.3345\n",
      "Epoch 37/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3541 - val_loss: 0.3843\n",
      "Epoch 38/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3534 - val_loss: 0.3329\n",
      "Epoch 39/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3524 - val_loss: 0.3933\n",
      "Epoch 40/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3522 - val_loss: 0.3314\n",
      "Epoch 41/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3510 - val_loss: 0.3985\n",
      "Epoch 42/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3507 - val_loss: 0.3299\n",
      "Epoch 43/100\n",
      "11610/11610 [==============================] - 0s 43us/sample - loss: 0.3498 - val_loss: 0.3466\n",
      "Epoch 44/100\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.3488 - val_loss: 0.3414\n",
      "Epoch 45/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3484 - val_loss: 0.3436\n",
      "Epoch 46/100\n",
      "11610/11610 [==============================] - ETA: 0s - loss: 0.347 - 0s 38us/sample - loss: 0.3479 - val_loss: 0.3289\n",
      "Epoch 47/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3468 - val_loss: 0.3655\n",
      "Epoch 48/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3467 - val_loss: 0.3285\n",
      "Epoch 49/100\n",
      "11610/11610 [==============================] - 0s 41us/sample - loss: 0.3461 - val_loss: 0.3964\n",
      "Epoch 50/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3455 - val_loss: 0.3299\n",
      "Epoch 51/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3451 - val_loss: 0.3349\n",
      "Epoch 52/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3446 - val_loss: 0.3328\n",
      "Epoch 53/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3440 - val_loss: 0.3580\n",
      "Epoch 54/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3433 - val_loss: 0.3316\n",
      "Epoch 55/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3429 - val_loss: 0.3252\n",
      "Epoch 56/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3422 - val_loss: 0.3857\n",
      "Epoch 57/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3422 - val_loss: 0.3237\n",
      "Epoch 58/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3414 - val_loss: 0.3387\n",
      "Epoch 59/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3409 - val_loss: 0.3226\n",
      "Epoch 60/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3403 - val_loss: 0.4004\n",
      "Epoch 61/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3406 - val_loss: 0.3220\n",
      "Epoch 62/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3396 - val_loss: 0.3940\n",
      "Epoch 63/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3393 - val_loss: 0.3282\n",
      "Epoch 64/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3391 - val_loss: 0.3584\n",
      "Epoch 65/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3383 - val_loss: 0.3217\n",
      "Epoch 66/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3380 - val_loss: 0.3738\n",
      "Epoch 67/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3378 - val_loss: 0.3397\n",
      "Epoch 68/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3373 - val_loss: 0.3194\n",
      "Epoch 69/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3367 - val_loss: 0.3785\n",
      "Epoch 70/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3368 - val_loss: 0.3205\n",
      "Epoch 71/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3358 - val_loss: 0.3817\n",
      "Epoch 72/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3358 - val_loss: 0.3350\n",
      "Epoch 73/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3350 - val_loss: 0.3332\n",
      "Epoch 74/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3350 - val_loss: 0.3179\n",
      "Epoch 75/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3345 - val_loss: 0.3476\n",
      "Epoch 76/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3341 - val_loss: 0.3196\n",
      "Epoch 77/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3338 - val_loss: 0.3608\n",
      "Epoch 78/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3336 - val_loss: 0.3295\n",
      "Epoch 79/100\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.3331 - val_loss: 0.3927\n",
      "Epoch 80/100\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.3332 - val_loss: 0.3167\n",
      "Epoch 81/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3317 - val_loss: 0.3783\n",
      "Epoch 82/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3322 - val_loss: 0.3154\n",
      "Epoch 83/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3315 - val_loss: 0.3538\n",
      "Epoch 84/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3312 - val_loss: 0.3282\n",
      "Epoch 85/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3307 - val_loss: 0.3379\n",
      "Epoch 86/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3306 - val_loss: 0.3163\n",
      "Epoch 87/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3302 - val_loss: 0.3377\n",
      "Epoch 88/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3300 - val_loss: 0.3340\n",
      "Epoch 89/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3295 - val_loss: 0.3158\n",
      "Epoch 90/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3294 - val_loss: 0.3519\n",
      "Epoch 91/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3289 - val_loss: 0.3142\n",
      "Epoch 92/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3286 - val_loss: 0.3619\n",
      "Epoch 93/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3282 - val_loss: 0.3270\n",
      "Epoch 94/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3282 - val_loss: 0.4632\n",
      "Epoch 95/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3284 - val_loss: 0.3371\n",
      "Epoch 96/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3275 - val_loss: 0.4659\n",
      "Epoch 97/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3278 - val_loss: 0.3156\n",
      "Epoch 98/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3265 - val_loss: 0.3259\n",
      "Epoch 99/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3262 - val_loss: 0.3407\n",
      "Epoch 100/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3264 - val_loss: 0.3176\n",
      "5160/5160 [==============================] - 0s 19us/sample - loss: 0.3271\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss=\"mse\", optimizer=keras.optimizers.SGD(lr=1e-3))\n",
    "early_stopping_cb = keras.callbacks.EarlyStopping(patience=10,\n",
    "                                                  restore_best_weights=True)\n",
    "history = model.fit(X_train, y_train, epochs=100,\n",
    "                    validation_data=(X_valid, y_valid),\n",
    "                    callbacks=[checkpoint_cb, early_stopping_cb])\n",
    "mse_test = model.evaluate(X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [],
   "source": [
    "class PrintValTrainRatioCallback(keras.callbacks.Callback):\n",
    "    def on_epoch_end(self, epoch, logs):\n",
    "        print(\"\\nval/train: {:.2f}\".format(logs[\"val_loss\"] / logs[\"loss\"]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 11610 samples, validate on 3870 samples\n",
      "10464/11610 [==========================>...] - ETA: 0s - loss: 0.3204\n",
      "val/train: 1.16\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3256 - val_loss: 0.3785\n"
     ]
    }
   ],
   "source": [
    "val_train_ratio_cb = PrintValTrainRatioCallback()\n",
    "history = model.fit(X_train, y_train, epochs=1,\n",
    "                    validation_data=(X_valid, y_valid),\n",
    "                    callbacks=[val_train_ratio_cb])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# TensorBoard"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "root_logdir = os.path.join(os.curdir, \"my_logs\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'.\\\\my_logs\\\\run_2020_11_10-12_30_31'"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def get_run_logdir():\n",
    "    import time\n",
    "    run_id = time.strftime(\"run_%Y_%m_%d-%H_%M_%S\")\n",
    "    return os.path.join(root_logdir, run_id)\n",
    "\n",
    "run_logdir = get_run_logdir()\n",
    "run_logdir"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "keras.backend.clear_session()\n",
    "np.random.seed(42)\n",
    "tf.random.set_seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.models.Sequential([\n",
    "    keras.layers.Dense(30, activation=\"relu\", input_shape=[8]),\n",
    "    keras.layers.Dense(30, activation=\"relu\"),\n",
    "    keras.layers.Dense(1)\n",
    "])    \n",
    "model.compile(loss=\"mse\", optimizer=keras.optimizers.SGD(lr=1e-3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 11610 samples, validate on 3870 samples\n",
      "Epoch 1/30\n",
      "11610/11610 [==============================] - 1s 62us/sample - loss: 1.8423 - val_loss: 5.2165\n",
      "Epoch 2/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.6876 - val_loss: 0.7732\n",
      "Epoch 3/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.5954 - val_loss: 0.5446\n",
      "Epoch 4/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.5553 - val_loss: 0.5425\n",
      "Epoch 5/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.5268 - val_loss: 0.5539\n",
      "Epoch 6/30\n",
      "11610/11610 [==============================] - 1s 43us/sample - loss: 0.5049 - val_loss: 0.4701\n",
      "Epoch 7/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.4852 - val_loss: 0.4562\n",
      "Epoch 8/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.4706 - val_loss: 0.4452\n",
      "Epoch 9/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.4576 - val_loss: 0.4406\n",
      "Epoch 10/30\n",
      "11610/11610 [==============================] - 1s 43us/sample - loss: 0.4476 - val_loss: 0.4185\n",
      "Epoch 11/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.4388 - val_loss: 0.4285\n",
      "Epoch 12/30\n",
      "11610/11610 [==============================] - 1s 48us/sample - loss: 0.4313 - val_loss: 0.4071\n",
      "Epoch 13/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.4252 - val_loss: 0.3998\n",
      "Epoch 14/30\n",
      "11610/11610 [==============================] - 1s 43us/sample - loss: 0.4198 - val_loss: 0.3970\n",
      "Epoch 15/30\n",
      "11610/11610 [==============================] - 1s 43us/sample - loss: 0.4152 - val_loss: 0.4115\n",
      "Epoch 16/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.4109 - val_loss: 0.3849\n",
      "Epoch 17/30\n",
      "11610/11610 [==============================] - 1s 43us/sample - loss: 0.4073 - val_loss: 0.3862\n",
      "Epoch 18/30\n",
      "11610/11610 [==============================] - 0s 43us/sample - loss: 0.4037 - val_loss: 0.3907\n",
      "Epoch 19/30\n",
      "11610/11610 [==============================] - 0s 43us/sample - loss: 0.4003 - val_loss: 0.3751\n",
      "Epoch 20/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.3974 - val_loss: 0.3711\n",
      "Epoch 21/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.3945 - val_loss: 0.3733\n",
      "Epoch 22/30\n",
      "11610/11610 [==============================] - 0s 43us/sample - loss: 0.3919 - val_loss: 0.3676\n",
      "Epoch 23/30\n",
      "11610/11610 [==============================] - 1s 43us/sample - loss: 0.3893 - val_loss: 0.3669\n",
      "Epoch 24/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.3869 - val_loss: 0.3614\n",
      "Epoch 25/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.3846 - val_loss: 0.3600\n",
      "Epoch 26/30\n",
      "11610/11610 [==============================] - 1s 48us/sample - loss: 0.3825 - val_loss: 0.3578\n",
      "Epoch 27/30\n",
      "11610/11610 [==============================] - 1s 46us/sample - loss: 0.3802 - val_loss: 0.3676\n",
      "Epoch 28/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.3781 - val_loss: 0.3545\n",
      "Epoch 29/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.3761 - val_loss: 0.3612\n",
      "Epoch 30/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.3742 - val_loss: 0.3555\n"
     ]
    }
   ],
   "source": [
    "tensorboard_cb = keras.callbacks.TensorBoard(run_logdir)\n",
    "history = model.fit(X_train, y_train, epochs=30,\n",
    "                    validation_data=(X_valid, y_valid),\n",
    "                    callbacks=[checkpoint_cb, tensorboard_cb])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To start the TensorBoard server, use terminal:\n",
    "```\n",
    "$ tensorboard --logdir=./my_logs --port=6006\n",
    "```\n",
    "Then, in a web browser, use localhost:6006. Once you are done, press Ctrl-C in the terminal to shutdown the TensorBoard server.\n",
    "\n",
    "Alternatively, use the Jupyter extension:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Reusing TensorBoard on port 6006 (pid 692), started 3:20:19 ago. (Use '!kill 692' to kill it.)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "\n",
       "      <iframe id=\"tensorboard-frame-e1ae5b0aed999208\" width=\"100%\" height=\"800\" frameborder=\"0\">\n",
       "      </iframe>\n",
       "      <script>\n",
       "        (function() {\n",
       "          const frame = document.getElementById(\"tensorboard-frame-e1ae5b0aed999208\");\n",
       "          const url = new URL(\"/\", window.location);\n",
       "          url.port = 6006;\n",
       "          frame.src = url;\n",
       "        })();\n",
       "      </script>\n",
       "  "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%load_ext tensorboard\n",
    "%tensorboard --logdir=./my_logs --port=6006"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'.\\\\my_logs\\\\run_2020_11_10-12_30_47'"
      ]
     },
     "execution_count": 79,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "run_logdir2 = get_run_logdir() # make a new TF-board callback with a different name for another model\n",
    "run_logdir2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    "keras.backend.clear_session()\n",
    "np.random.seed(42)\n",
    "tf.random.set_seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.models.Sequential([\n",
    "    keras.layers.Dense(30, activation=\"relu\", input_shape=[8]),\n",
    "    keras.layers.Dense(30, activation=\"relu\"),\n",
    "    keras.layers.Dense(1)\n",
    "])    \n",
    "model.compile(loss=\"mse\", optimizer=keras.optimizers.SGD(lr=0.05))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 11610 samples, validate on 3870 samples\n",
      "Epoch 1/30\n",
      "11610/11610 [==============================] - 1s 73us/sample - loss: 5.6341 - val_loss: 1.3205\n",
      "Epoch 2/30\n",
      "11610/11610 [==============================] - 1s 49us/sample - loss: 1.2704 - val_loss: 1.0757\n",
      "Epoch 3/30\n",
      "11610/11610 [==============================] - 1s 47us/sample - loss: 0.9370 - val_loss: 0.7769\n",
      "Epoch 4/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.8854 - val_loss: 0.8254\n",
      "Epoch 5/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.9501 - val_loss: 0.9415\n",
      "Epoch 6/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.8283 - val_loss: 0.6111\n",
      "Epoch 7/30\n",
      "11610/11610 [==============================] - 1s 47us/sample - loss: 0.8095 - val_loss: 0.7394\n",
      "Epoch 8/30\n",
      "11610/11610 [==============================] - 1s 46us/sample - loss: 1.1857 - val_loss: 1.0356\n",
      "Epoch 9/30\n",
      "11610/11610 [==============================] - 1s 47us/sample - loss: 0.8920 - val_loss: 0.6564\n",
      "Epoch 10/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 1.2295 - val_loss: 1.1277\n",
      "Epoch 11/30\n",
      "11610/11610 [==============================] - 1s 48us/sample - loss: 0.8992 - val_loss: 0.6219\n",
      "Epoch 12/30\n",
      "11610/11610 [==============================] - 1s 50us/sample - loss: 0.7697 - val_loss: 0.6959\n",
      "Epoch 13/30\n",
      "11610/11610 [==============================] - 1s 48us/sample - loss: 0.7375 - val_loss: 0.6471\n",
      "Epoch 14/30\n",
      "11610/11610 [==============================] - 1s 47us/sample - loss: 0.9825 - val_loss: 1.0278\n",
      "Epoch 15/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.9653 - val_loss: 0.9092\n",
      "Epoch 16/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.9008 - val_loss: 0.8451\n",
      "Epoch 17/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.7838 - val_loss: 0.5818\n",
      "Epoch 18/30\n",
      "11610/11610 [==============================] - 1s 46us/sample - loss: 0.9960 - val_loss: 1.0598\n",
      "Epoch 19/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.7887 - val_loss: 0.6405\n",
      "Epoch 20/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.5706 - val_loss: 0.5093\n",
      "Epoch 21/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.5192 - val_loss: 0.5457\n",
      "Epoch 22/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.4971 - val_loss: 0.4644\n",
      "Epoch 23/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.4815 - val_loss: 0.4120\n",
      "Epoch 24/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.4678 - val_loss: 0.4853\n",
      "Epoch 25/30\n",
      "11610/11610 [==============================] - 1s 43us/sample - loss: 0.4490 - val_loss: 0.3972\n",
      "Epoch 26/30\n",
      "11610/11610 [==============================] - 1s 45us/sample - loss: 0.4348 - val_loss: 0.4258\n",
      "Epoch 27/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.4377 - val_loss: 0.3829\n",
      "Epoch 28/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.4353 - val_loss: 0.3860\n",
      "Epoch 29/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.4209 - val_loss: 0.3763\n",
      "Epoch 30/30\n",
      "11610/11610 [==============================] - 1s 44us/sample - loss: 0.4441 - val_loss: 0.4186\n"
     ]
    }
   ],
   "source": [
    "tensorboard_cb = keras.callbacks.TensorBoard(run_logdir2)\n",
    "history = model.fit(X_train, y_train, epochs=30,\n",
    "                    validation_data=(X_valid, y_valid),\n",
    "                    callbacks=[checkpoint_cb, tensorboard_cb])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Hyperparameter Tuning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [],
   "source": [
    "keras.backend.clear_session()\n",
    "np.random.seed(42)\n",
    "tf.random.set_seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [],
   "source": [
    "def build_model(n_hidden=1, n_neurons=30, learning_rate=3e-3, input_shape=[8]):\n",
    "    model = keras.models.Sequential()\n",
    "    model.add(keras.layers.InputLayer(input_shape=input_shape))\n",
    "    for layer in range(n_hidden):\n",
    "        model.add(keras.layers.Dense(n_neurons, activation=\"relu\"))\n",
    "    model.add(keras.layers.Dense(1))\n",
    "    optimizer = keras.optimizers.SGD(lr=learning_rate)\n",
    "    model.compile(loss=\"mse\", optimizer=optimizer)\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [],
   "source": [
    "keras_reg = keras.wrappers.scikit_learn.KerasRegressor(build_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 11610 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "11610/11610 [==============================] - 1s 54us/sample - loss: 1.1399 - val_loss: 24.4309\n",
      "Epoch 2/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.7393 - val_loss: 3.2896\n",
      "Epoch 3/100\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.5374 - val_loss: 0.6080\n",
      "Epoch 4/100\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.4796 - val_loss: 0.4532\n",
      "Epoch 5/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.4536 - val_loss: 0.4221\n",
      "Epoch 6/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.4373 - val_loss: 0.4144\n",
      "Epoch 7/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.4256 - val_loss: 0.4025\n",
      "Epoch 8/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.4178 - val_loss: 0.3938\n",
      "Epoch 9/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.4114 - val_loss: 0.4085\n",
      "Epoch 10/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.4060 - val_loss: 0.3974\n",
      "Epoch 11/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.4018 - val_loss: 0.3847\n",
      "Epoch 12/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3978 - val_loss: 0.3818\n",
      "Epoch 13/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3943 - val_loss: 0.3820\n",
      "Epoch 14/100\n",
      "11610/11610 [==============================] - 0s 41us/sample - loss: 0.3915 - val_loss: 0.3766\n",
      "Epoch 15/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3888 - val_loss: 0.4268\n",
      "Epoch 16/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3862 - val_loss: 0.3638\n",
      "Epoch 17/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3842 - val_loss: 0.3666\n",
      "Epoch 18/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3815 - val_loss: 0.3948\n",
      "Epoch 19/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3797 - val_loss: 0.3589\n",
      "Epoch 20/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3777 - val_loss: 0.4058\n",
      "Epoch 21/100\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.3763 - val_loss: 0.3662\n",
      "Epoch 22/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3742 - val_loss: 0.4100\n",
      "Epoch 23/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3729 - val_loss: 0.3522\n",
      "Epoch 24/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3718 - val_loss: 0.3518\n",
      "Epoch 25/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3701 - val_loss: 0.3586\n",
      "Epoch 26/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3689 - val_loss: 0.3600\n",
      "Epoch 27/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3674 - val_loss: 0.4042\n",
      "Epoch 28/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3666 - val_loss: 0.3520\n",
      "Epoch 29/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3649 - val_loss: 0.4091\n",
      "Epoch 30/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3645 - val_loss: 0.3916\n",
      "Epoch 31/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3636 - val_loss: 0.3426\n",
      "Epoch 32/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3626 - val_loss: 0.3516\n",
      "Epoch 33/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3616 - val_loss: 0.3830\n",
      "Epoch 34/100\n",
      "11610/11610 [==============================] - 0s 42us/sample - loss: 0.3613 - val_loss: 0.3411\n",
      "Epoch 35/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3600 - val_loss: 0.3656\n",
      "Epoch 36/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3593 - val_loss: 0.3402\n",
      "Epoch 37/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3583 - val_loss: 0.3949\n",
      "Epoch 38/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3578 - val_loss: 0.3376\n",
      "Epoch 39/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3563 - val_loss: 0.4037\n",
      "Epoch 40/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3567 - val_loss: 0.3391\n",
      "Epoch 41/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3553 - val_loss: 0.4145\n",
      "Epoch 42/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3552 - val_loss: 0.3376\n",
      "Epoch 43/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3539 - val_loss: 0.3485\n",
      "Epoch 44/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3533 - val_loss: 0.3355\n",
      "Epoch 45/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3529 - val_loss: 0.3329\n",
      "Epoch 46/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3522 - val_loss: 0.3393\n",
      "Epoch 47/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3510 - val_loss: 0.4172\n",
      "Epoch 48/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3515 - val_loss: 0.3758\n",
      "Epoch 49/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3506 - val_loss: 0.4285\n",
      "Epoch 50/100\n",
      "11610/11610 [==============================] - 0s 39us/sample - loss: 0.3506 - val_loss: 0.4004\n",
      "Epoch 51/100\n",
      "11610/11610 [==============================] - 0s 40us/sample - loss: 0.3501 - val_loss: 0.3300\n",
      "Epoch 52/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3495 - val_loss: 0.3300\n",
      "Epoch 53/100\n",
      "11610/11610 [==============================] - 0s 35us/sample - loss: 0.3485 - val_loss: 0.3599\n",
      "Epoch 54/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3479 - val_loss: 0.3424\n",
      "Epoch 55/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3475 - val_loss: 0.3752\n",
      "Epoch 56/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3472 - val_loss: 0.4231\n",
      "Epoch 57/100\n",
      "11610/11610 [==============================] - 0s 36us/sample - loss: 0.3474 - val_loss: 0.3728\n",
      "Epoch 58/100\n",
      "11610/11610 [==============================] - 0s 35us/sample - loss: 0.3464 - val_loss: 0.3455\n",
      "Epoch 59/100\n",
      "11610/11610 [==============================] - 0s 38us/sample - loss: 0.3458 - val_loss: 0.3647\n",
      "Epoch 60/100\n",
      "11610/11610 [==============================] - 0s 37us/sample - loss: 0.3451 - val_loss: 0.4226\n",
      "Epoch 61/100\n",
      "11610/11610 [==============================] - 0s 41us/sample - loss: 0.3456 - val_loss: 0.3635\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x1c6161cae48>"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "keras_reg.fit(X_train, y_train, epochs=100,\n",
    "              validation_data=(X_valid, y_valid),\n",
    "              callbacks=[keras.callbacks.EarlyStopping(patience=10)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5160/5160 [==============================] - 0s 21us/sample - loss: 0.3464\n"
     ]
    }
   ],
   "source": [
    "mse_test = keras_reg.score(X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = keras_reg.predict(X_new)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(42)\n",
    "tf.random.set_seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 10 candidates, totalling 30 fits\n",
      "[CV] learning_rate=0.001683454924600351, n_hidden=0, n_neurons=15 ....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 59us/sample - loss: 3.5574 - val_loss: 1.8536\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 37us/sample - loss: 1.3316 - val_loss: 0.9380\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.8573 - val_loss: 0.8545\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.7344 - val_loss: 0.9545\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.6943 - val_loss: 0.7248\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.6682 - val_loss: 0.7356\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.6494 - val_loss: 0.9732\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6408 - val_loss: 0.6175\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.6256 - val_loss: 0.5877\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6149 - val_loss: 0.6164\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6058 - val_loss: 0.5851\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5961 - val_loss: 0.7040\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5905 - val_loss: 0.5594\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5797 - val_loss: 0.8668\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5764 - val_loss: 0.9031\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5745 - val_loss: 0.5583\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5635 - val_loss: 0.9090\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5639 - val_loss: 0.7540\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5608 - val_loss: 0.5828\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5562 - val_loss: 0.5237\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5528 - val_loss: 0.5610\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5497 - val_loss: 0.6308\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5466 - val_loss: 0.7470\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5477 - val_loss: 0.5209\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5416 - val_loss: 0.7251\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5432 - val_loss: 0.6298\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5390 - val_loss: 0.8002\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5417 - val_loss: 0.5708\n",
      "Epoch 29/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5380 - val_loss: 0.5888\n",
      "Epoch 30/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5356 - val_loss: 0.7386\n",
      "Epoch 31/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5382 - val_loss: 0.5329\n",
      "Epoch 32/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5334 - val_loss: 0.7252\n",
      "Epoch 33/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5364 - val_loss: 0.5224\n",
      "Epoch 34/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5339 - val_loss: 0.5262\n",
      "3870/3870 [==============================] - 0s 21us/sample - loss: 0.5463\n",
      "[CV]  learning_rate=0.001683454924600351, n_hidden=0, n_neurons=15, total=  10.8s\n",
      "[CV] learning_rate=0.001683454924600351, n_hidden=0, n_neurons=15 ....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed:   10.7s remaining:    0.0s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 1s 66us/sample - loss: 3.5541 - val_loss: 23.2063\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 37us/sample - loss: 1.4756 - val_loss: 10.9096\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 1.0159 - val_loss: 4.5194\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.8738 - val_loss: 1.5793\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.8034 - val_loss: 0.7218\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.7549 - val_loss: 1.1511\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.7166 - val_loss: 2.3325\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.6854 - val_loss: 3.9493\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.6593 - val_loss: 5.7393\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6375 - val_loss: 7.6052\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.6190 - val_loss: 9.3527\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6033 - val_loss: 11.0010\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5898 - val_loss: 12.6250\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5783 - val_loss: 13.9778\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5684 - val_loss: 15.1585\n",
      "3870/3870 [==============================] - 0s 20us/sample - loss: 0.9136\n",
      "[CV]  learning_rate=0.001683454924600351, n_hidden=0, n_neurons=15, total=   4.9s\n",
      "[CV] learning_rate=0.001683454924600351, n_hidden=0, n_neurons=15 ....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 80us/sample - loss: 3.2309 - val_loss: 1.4092\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.9653 - val_loss: 0.7052\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.6144 - val_loss: 0.7331\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5510 - val_loss: 0.5728\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5343 - val_loss: 0.7233\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5342 - val_loss: 0.6482\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5322 - val_loss: 0.6808\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.5310 - val_loss: 0.7642\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5310 - val_loss: 0.8269\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5318 - val_loss: 0.8031\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5335 - val_loss: 0.5911\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5310 - val_loss: 0.6315\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5319 - val_loss: 0.4982\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5308 - val_loss: 0.5125\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5309 - val_loss: 0.4935\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5303 - val_loss: 0.5501\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5289 - val_loss: 0.6833\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5320 - val_loss: 0.5203\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5297 - val_loss: 0.5818\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5290 - val_loss: 0.6962\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5309 - val_loss: 0.6175\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5303 - val_loss: 0.6093\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5301 - val_loss: 0.6215\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5308 - val_loss: 0.5782\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5302 - val_loss: 0.5895\n",
      "3870/3870 [==============================] - 0s 18us/sample - loss: 0.5308\n",
      "[CV]  learning_rate=0.001683454924600351, n_hidden=0, n_neurons=15, total=   8.2s\n",
      "[CV] learning_rate=0.008731907739399206, n_hidden=0, n_neurons=21 ....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 57us/sample - loss: 1.4568 - val_loss: 10.1894\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6432 - val_loss: 0.8423\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5927 - val_loss: 0.5931\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5738 - val_loss: 0.9545\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.6437 - val_loss: 26.9553\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.6352 - val_loss: 49.8458\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 2.0469 - val_loss: 161.2909\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 1.0236 - val_loss: 296.4855\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 2.0421 - val_loss: 1833.8984\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 10.1616 - val_loss: 6133.6058\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 37.8901 - val_loss: 26555.2970\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 252.6434 - val_loss: 101521.9690\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 492.7466 - val_loss: 415890.0651\n",
      "3870/3870 [==============================] - 0s 18us/sample - loss: 1144.4789\n",
      "[CV]  learning_rate=0.008731907739399206, n_hidden=0, n_neurons=21, total=   4.3s\n",
      "[CV] learning_rate=0.008731907739399206, n_hidden=0, n_neurons=21 ....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 58us/sample - loss: 1.1508 - val_loss: 21.3931\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5340 - val_loss: 21.6021\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5226 - val_loss: 21.4140\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5148 - val_loss: 19.0061\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5125 - val_loss: 18.3419\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5082 - val_loss: 18.0573\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5050 - val_loss: 16.9085\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5058 - val_loss: 17.2824\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5057 - val_loss: 16.9900\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5035 - val_loss: 18.3345\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5045 - val_loss: 16.7249\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5023 - val_loss: 16.9628\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5023 - val_loss: 19.5300\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 37us/sample - loss: 0.5022 - val_loss: 18.6929\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5031 - val_loss: 17.7248\n",
      "Epoch 16/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 37us/sample - loss: 0.5027 - val_loss: 20.9402\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5022 - val_loss: 20.5891\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5031 - val_loss: 20.7123\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5014 - val_loss: 17.3551\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5013 - val_loss: 20.4442\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5037 - val_loss: 20.7221\n",
      "3870/3870 [==============================] - 0s 18us/sample - loss: 1.0074\n",
      "[CV]  learning_rate=0.008731907739399206, n_hidden=0, n_neurons=21, total=   6.7s\n",
      "[CV] learning_rate=0.008731907739399206, n_hidden=0, n_neurons=21 ....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 60us/sample - loss: 1.4056 - val_loss: 28.7777\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.9238 - val_loss: 125.2706\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 2.8216 - val_loss: 268.4825\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 4.9607 - val_loss: 622.8712\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 11.6930 - val_loss: 1247.5503\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 16.1341 - val_loss: 2481.0882\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 48.1242 - val_loss: 4792.8348\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 119.1232 - val_loss: 8994.9969\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 266.2211 - val_loss: 17774.0706\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 489.5452 - val_loss: 33844.4918\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 37us/sample - loss: 370.8847 - val_loss: 64184.8462\n",
      "3870/3870 [==============================] - 0s 18us/sample - loss: 65.7429\n",
      "[CV]  learning_rate=0.008731907739399206, n_hidden=0, n_neurons=21, total=   3.6s\n",
      "[CV] learning_rate=0.0006154014789262348, n_hidden=2, n_neurons=87 ...\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 76us/sample - loss: 2.4669 - val_loss: 5.0606\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 1.1144 - val_loss: 1.4034\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.8083 - val_loss: 0.7548\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.7204 - val_loss: 0.6779\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6790 - val_loss: 0.6473\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6502 - val_loss: 0.6205\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6261 - val_loss: 0.6041\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6055 - val_loss: 0.5773\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.5870 - val_loss: 0.5580\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5703 - val_loss: 0.5421\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5549 - val_loss: 0.5271\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 50us/sample - loss: 0.5409 - val_loss: 0.5152\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 49us/sample - loss: 0.5281 - val_loss: 0.5013\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5158 - val_loss: 0.4916\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.5054 - val_loss: 0.4796\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4950 - val_loss: 0.4721\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4859 - val_loss: 0.4610\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4775 - val_loss: 0.4528\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.4697 - val_loss: 0.4477\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4622 - val_loss: 0.4394\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4557 - val_loss: 0.4368\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4498 - val_loss: 0.4277\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4440 - val_loss: 0.4261\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4388 - val_loss: 0.4190\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4339 - val_loss: 0.4201\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4296 - val_loss: 0.4125\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4254 - val_loss: 0.4095\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4218 - val_loss: 0.4107\n",
      "Epoch 29/100\n",
      "7740/7740 [==============================] - 0s 50us/sample - loss: 0.4183 - val_loss: 0.4035\n",
      "Epoch 30/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4150 - val_loss: 0.4046\n",
      "Epoch 31/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4119 - val_loss: 0.4032\n",
      "Epoch 32/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4089 - val_loss: 0.4043\n",
      "Epoch 33/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4062 - val_loss: 0.3991\n",
      "Epoch 34/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4036 - val_loss: 0.4024\n",
      "Epoch 35/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4012 - val_loss: 0.3986\n",
      "Epoch 36/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3985 - val_loss: 0.3892\n",
      "Epoch 37/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.3968 - val_loss: 0.3883\n",
      "Epoch 38/100\n",
      "7740/7740 [==============================] - 0s 49us/sample - loss: 0.3945 - val_loss: 0.3954\n",
      "Epoch 39/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3927 - val_loss: 0.3982\n",
      "Epoch 40/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3908 - val_loss: 0.3901\n",
      "Epoch 41/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3891 - val_loss: 0.3864\n",
      "Epoch 42/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3873 - val_loss: 0.3872\n",
      "Epoch 43/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3856 - val_loss: 0.3946\n",
      "Epoch 44/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3840 - val_loss: 0.3847\n",
      "Epoch 45/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3825 - val_loss: 0.3998\n",
      "Epoch 46/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.3811 - val_loss: 0.3924\n",
      "Epoch 47/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3797 - val_loss: 0.4011\n",
      "Epoch 48/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3781 - val_loss: 0.3934\n",
      "Epoch 49/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3770 - val_loss: 0.3778\n",
      "Epoch 50/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3759 - val_loss: 0.3736\n",
      "Epoch 51/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3746 - val_loss: 0.3914\n",
      "Epoch 52/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3734 - val_loss: 0.3860\n",
      "Epoch 53/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.3722 - val_loss: 0.3707\n",
      "Epoch 54/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3713 - val_loss: 0.3751\n",
      "Epoch 55/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.3700 - val_loss: 0.3698\n",
      "Epoch 56/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3691 - val_loss: 0.3854\n",
      "Epoch 57/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3682 - val_loss: 0.3874\n",
      "Epoch 58/100\n",
      "7740/7740 [==============================] - 0s 50us/sample - loss: 0.3670 - val_loss: 0.3942\n",
      "Epoch 59/100\n",
      "7740/7740 [==============================] - 0s 51us/sample - loss: 0.3663 - val_loss: 0.3821\n",
      "Epoch 60/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3654 - val_loss: 0.3882\n",
      "Epoch 61/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3644 - val_loss: 0.3779\n",
      "Epoch 62/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.3634 - val_loss: 0.3693\n",
      "Epoch 63/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3628 - val_loss: 0.3734\n",
      "Epoch 64/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.3619 - val_loss: 0.3695\n",
      "Epoch 65/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3610 - val_loss: 0.3745\n",
      "Epoch 66/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3603 - val_loss: 0.3764\n",
      "Epoch 67/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3596 - val_loss: 0.3654\n",
      "Epoch 68/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3588 - val_loss: 0.3681\n",
      "Epoch 69/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3579 - val_loss: 0.3829\n",
      "Epoch 70/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3574 - val_loss: 0.3662\n",
      "Epoch 71/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3565 - val_loss: 0.3689\n",
      "Epoch 72/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3560 - val_loss: 0.3616\n",
      "Epoch 73/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3553 - val_loss: 0.3783\n",
      "Epoch 74/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3544 - val_loss: 0.3938\n",
      "Epoch 75/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3541 - val_loss: 0.3822\n",
      "Epoch 76/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3535 - val_loss: 0.3626\n",
      "Epoch 77/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3528 - val_loss: 0.3579\n",
      "Epoch 78/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3520 - val_loss: 0.3700\n",
      "Epoch 79/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3515 - val_loss: 0.3780\n",
      "Epoch 80/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3509 - val_loss: 0.3747\n",
      "Epoch 81/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3504 - val_loss: 0.3704\n",
      "Epoch 82/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3499 - val_loss: 0.3727\n",
      "Epoch 83/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3493 - val_loss: 0.3852\n",
      "Epoch 84/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3488 - val_loss: 0.3726\n",
      "Epoch 85/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3482 - val_loss: 0.3575\n",
      "Epoch 86/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3476 - val_loss: 0.3785\n",
      "Epoch 87/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3469 - val_loss: 0.3832\n",
      "Epoch 88/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3468 - val_loss: 0.3625\n",
      "Epoch 89/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3462 - val_loss: 0.3652\n",
      "Epoch 90/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3456 - val_loss: 0.3666\n",
      "Epoch 91/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3453 - val_loss: 0.3603\n",
      "Epoch 92/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3447 - val_loss: 0.3770\n",
      "Epoch 93/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3443 - val_loss: 0.3721\n",
      "Epoch 94/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3439 - val_loss: 0.3569\n",
      "Epoch 95/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3432 - val_loss: 0.3556\n",
      "Epoch 96/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3429 - val_loss: 0.3712\n",
      "Epoch 97/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3426 - val_loss: 0.3563\n",
      "Epoch 98/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.3420 - val_loss: 0.3756\n",
      "Epoch 99/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3413 - val_loss: 0.3725\n",
      "Epoch 100/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3413 - val_loss: 0.3547\n",
      "3870/3870 [==============================] - 0s 21us/sample - loss: 0.3647\n",
      "[CV]  learning_rate=0.0006154014789262348, n_hidden=2, n_neurons=87, total=  34.9s\n",
      "[CV] learning_rate=0.0006154014789262348, n_hidden=2, n_neurons=87 ...\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 67us/sample - loss: 2.7764 - val_loss: 17.7273\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 1.1021 - val_loss: 14.8833\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.8031 - val_loss: 10.7757\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 49us/sample - loss: 0.7045 - val_loss: 7.9756\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.6578 - val_loss: 6.0529\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6260 - val_loss: 4.7215\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6005 - val_loss: 3.5591\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5789 - val_loss: 2.8484\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5598 - val_loss: 2.1834\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5433 - val_loss: 1.8093\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5284 - val_loss: 1.4123\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5151 - val_loss: 1.1408\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5037 - val_loss: 0.9795\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4931 - val_loss: 0.8257\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4838 - val_loss: 0.7039\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4755 - val_loss: 0.6402\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4679 - val_loss: 0.5641\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4611 - val_loss: 0.5326\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4550 - val_loss: 0.4901\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4492 - val_loss: 0.4640\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4443 - val_loss: 0.4593\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4396 - val_loss: 0.4449\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4353 - val_loss: 0.4364\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4313 - val_loss: 0.4345\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4276 - val_loss: 0.4275\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4240 - val_loss: 0.4316\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4209 - val_loss: 0.4256\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4180 - val_loss: 0.4302\n",
      "Epoch 29/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4152 - val_loss: 0.4260\n",
      "Epoch 30/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4125 - val_loss: 0.4350\n",
      "Epoch 31/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4099 - val_loss: 0.4413\n",
      "Epoch 32/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4076 - val_loss: 0.4439\n",
      "Epoch 33/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4052 - val_loss: 0.4513\n",
      "Epoch 34/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4029 - val_loss: 0.4671\n",
      "Epoch 35/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4010 - val_loss: 0.4686\n",
      "Epoch 36/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3988 - val_loss: 0.4745\n",
      "Epoch 37/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3971 - val_loss: 0.4899\n",
      "3870/3870 [==============================] - 0s 32us/sample - loss: 0.4098\n",
      "[CV]  learning_rate=0.0006154014789262348, n_hidden=2, n_neurons=87, total=  13.1s\n",
      "[CV] learning_rate=0.0006154014789262348, n_hidden=2, n_neurons=87 ...\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 67us/sample - loss: 2.8244 - val_loss: 4.9928\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 1.1397 - val_loss: 1.1758\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.8399 - val_loss: 0.8181\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.7621 - val_loss: 0.7366\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.7220 - val_loss: 0.6865\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6927 - val_loss: 0.6574\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6684 - val_loss: 0.6344\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6467 - val_loss: 0.6140\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6270 - val_loss: 0.5933\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6087 - val_loss: 0.5820\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5914 - val_loss: 0.5628\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.5756 - val_loss: 0.5565\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.5606 - val_loss: 0.5512\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5470 - val_loss: 0.5477\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.5345 - val_loss: 0.5238\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.5229 - val_loss: 0.5164\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.5122 - val_loss: 0.4974\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.5021 - val_loss: 0.4972\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.4932 - val_loss: 0.4757\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4848 - val_loss: 0.4819\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4772 - val_loss: 0.4767\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.4702 - val_loss: 0.4618\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4636 - val_loss: 0.4468\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.4577 - val_loss: 0.4685\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.4521 - val_loss: 0.4504\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4470 - val_loss: 0.4555\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4423 - val_loss: 0.4597\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4379 - val_loss: 0.4511\n",
      "Epoch 29/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4338 - val_loss: 0.4288\n",
      "Epoch 30/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4300 - val_loss: 0.4415\n",
      "Epoch 31/100\n",
      "7740/7740 [==============================] - 0s 51us/sample - loss: 0.4265 - val_loss: 0.4289\n",
      "Epoch 32/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4233 - val_loss: 0.4365\n",
      "Epoch 33/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.4202 - val_loss: 0.4348\n",
      "Epoch 34/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.4172 - val_loss: 0.4616\n",
      "Epoch 35/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.4149 - val_loss: 0.4292\n",
      "Epoch 36/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.4123 - val_loss: 0.4169\n",
      "Epoch 37/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4099 - val_loss: 0.4306\n",
      "Epoch 38/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4079 - val_loss: 0.4425\n",
      "Epoch 39/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4057 - val_loss: 0.4601\n",
      "Epoch 40/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4039 - val_loss: 0.4113\n",
      "Epoch 41/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4019 - val_loss: 0.4174\n",
      "Epoch 42/100\n",
      "7740/7740 [==============================] - 0s 52us/sample - loss: 0.4000 - val_loss: 0.4290\n",
      "Epoch 43/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3983 - val_loss: 0.4090\n",
      "Epoch 44/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3965 - val_loss: 0.4463\n",
      "Epoch 45/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3951 - val_loss: 0.4082\n",
      "Epoch 46/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3935 - val_loss: 0.4151\n",
      "Epoch 47/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3921 - val_loss: 0.4175\n",
      "Epoch 48/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3909 - val_loss: 0.4151\n",
      "Epoch 49/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3895 - val_loss: 0.4195\n",
      "Epoch 50/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3883 - val_loss: 0.3947\n",
      "Epoch 51/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3870 - val_loss: 0.3989\n",
      "Epoch 52/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3860 - val_loss: 0.3990\n",
      "Epoch 53/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3849 - val_loss: 0.3995\n",
      "Epoch 54/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3837 - val_loss: 0.3998\n",
      "Epoch 55/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3826 - val_loss: 0.4360\n",
      "Epoch 56/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.3817 - val_loss: 0.4396\n",
      "Epoch 57/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.3807 - val_loss: 0.4018\n",
      "Epoch 58/100\n",
      "7740/7740 [==============================] - 0s 49us/sample - loss: 0.3797 - val_loss: 0.4392\n",
      "Epoch 59/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3789 - val_loss: 0.4291\n",
      "Epoch 60/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3781 - val_loss: 0.4254\n",
      "3870/3870 [==============================] - 0s 24us/sample - loss: 0.3767\n",
      "[CV]  learning_rate=0.0006154014789262348, n_hidden=2, n_neurons=87, total=  21.4s\n",
      "[CV] learning_rate=0.0003920021771415983, n_hidden=3, n_neurons=24 ...\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 70us/sample - loss: 2.3872 - val_loss: 13.9258\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 1.2068 - val_loss: 6.0674\n",
      "Epoch 3/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.8942 - val_loss: 4.2789\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.8105 - val_loss: 3.3544\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.7686 - val_loss: 2.6742\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.7406 - val_loss: 2.2971\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.7187 - val_loss: 2.0216\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.7000 - val_loss: 1.7849\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.6825 - val_loss: 1.5810\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6676 - val_loss: 1.4502\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.6530 - val_loss: 1.3239\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.6398 - val_loss: 1.2281\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6268 - val_loss: 1.1328\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.6144 - val_loss: 1.0549\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6030 - val_loss: 0.9907\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5915 - val_loss: 0.9391\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5809 - val_loss: 0.8841\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5706 - val_loss: 0.8441\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5607 - val_loss: 0.8049\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5511 - val_loss: 0.7653\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5420 - val_loss: 0.7430\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5336 - val_loss: 0.7090\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5251 - val_loss: 0.6863\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5170 - val_loss: 0.6606\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5093 - val_loss: 0.6392\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5021 - val_loss: 0.6219\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4951 - val_loss: 0.6045\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4886 - val_loss: 0.5937\n",
      "Epoch 29/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4824 - val_loss: 0.5804\n",
      "Epoch 30/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4765 - val_loss: 0.5690\n",
      "Epoch 31/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4709 - val_loss: 0.5579\n",
      "Epoch 32/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4656 - val_loss: 0.5460\n",
      "Epoch 33/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4606 - val_loss: 0.5374\n",
      "Epoch 34/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4559 - val_loss: 0.5302\n",
      "Epoch 35/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4515 - val_loss: 0.5217\n",
      "Epoch 36/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4469 - val_loss: 0.5129\n",
      "Epoch 37/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4435 - val_loss: 0.5092\n",
      "Epoch 38/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4397 - val_loss: 0.5048\n",
      "Epoch 39/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4362 - val_loss: 0.4998\n",
      "Epoch 40/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4330 - val_loss: 0.4964\n",
      "Epoch 41/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4301 - val_loss: 0.4919\n",
      "Epoch 42/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4271 - val_loss: 0.4881\n",
      "Epoch 43/100\n",
      "7740/7740 [==============================] - 0s 49us/sample - loss: 0.4243 - val_loss: 0.4855\n",
      "Epoch 44/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.4217 - val_loss: 0.4819\n",
      "Epoch 45/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4193 - val_loss: 0.4816\n",
      "Epoch 46/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4171 - val_loss: 0.4787\n",
      "Epoch 47/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4147 - val_loss: 0.4792\n",
      "Epoch 48/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4126 - val_loss: 0.4752\n",
      "Epoch 49/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4105 - val_loss: 0.4720\n",
      "Epoch 50/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4088 - val_loss: 0.4707\n",
      "Epoch 51/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4071 - val_loss: 0.4716\n",
      "Epoch 52/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4053 - val_loss: 0.4717\n",
      "Epoch 53/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4037 - val_loss: 0.4692\n",
      "Epoch 54/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4022 - val_loss: 0.4693\n",
      "Epoch 55/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4005 - val_loss: 0.4673\n",
      "Epoch 56/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3993 - val_loss: 0.4693\n",
      "Epoch 57/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3979 - val_loss: 0.4681\n",
      "Epoch 58/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3964 - val_loss: 0.4700\n",
      "Epoch 59/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3953 - val_loss: 0.4698\n",
      "Epoch 60/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3941 - val_loss: 0.4687\n",
      "Epoch 61/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3928 - val_loss: 0.4699\n",
      "Epoch 62/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3915 - val_loss: 0.4665\n",
      "Epoch 63/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.3908 - val_loss: 0.4677\n",
      "Epoch 64/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3896 - val_loss: 0.4700\n",
      "Epoch 65/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3887 - val_loss: 0.4674\n",
      "Epoch 66/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3878 - val_loss: 0.4681\n",
      "Epoch 67/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3868 - val_loss: 0.4682\n",
      "Epoch 68/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3859 - val_loss: 0.4698\n",
      "Epoch 69/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3850 - val_loss: 0.4696\n",
      "Epoch 70/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3842 - val_loss: 0.4680\n",
      "Epoch 71/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3832 - val_loss: 0.4710\n",
      "Epoch 72/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3826 - val_loss: 0.4682\n",
      "3870/3870 [==============================] - 0s 22us/sample - loss: 0.3982\n",
      "[CV]  learning_rate=0.0003920021771415983, n_hidden=3, n_neurons=24, total=  24.9s\n",
      "[CV] learning_rate=0.0003920021771415983, n_hidden=3, n_neurons=24 ...\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 70us/sample - loss: 3.7649 - val_loss: 27.7302\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 2.0518 - val_loss: 41.6834\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 1.6095 - val_loss: 36.3809\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 1.3573 - val_loss: 27.7242\n",
      "Epoch 5/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 43us/sample - loss: 1.1658 - val_loss: 20.2997\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 1.0236 - val_loss: 14.4998\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.9247 - val_loss: 10.3206\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.8582 - val_loss: 7.2174\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.8127 - val_loss: 5.0609\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.7805 - val_loss: 3.8504\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7553 - val_loss: 2.6823\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7345 - val_loss: 1.9608\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7165 - val_loss: 1.5395\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7000 - val_loss: 1.1988\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.6849 - val_loss: 0.9615\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6712 - val_loss: 0.8662\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.6581 - val_loss: 0.7126\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 49us/sample - loss: 0.6455 - val_loss: 0.6655\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6340 - val_loss: 0.5995\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6223 - val_loss: 0.5767\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6115 - val_loss: 0.5779\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6009 - val_loss: 0.6051\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5909 - val_loss: 0.6417\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5814 - val_loss: 0.6742\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5723 - val_loss: 0.7591\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5635 - val_loss: 0.7764\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5552 - val_loss: 0.8860\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5474 - val_loss: 0.9032\n",
      "Epoch 29/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5401 - val_loss: 0.9833\n",
      "Epoch 30/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5329 - val_loss: 0.9995\n",
      "3870/3870 [==============================] - 0s 21us/sample - loss: 0.5486\n",
      "[CV]  learning_rate=0.0003920021771415983, n_hidden=3, n_neurons=24, total=  10.5s\n",
      "[CV] learning_rate=0.0003920021771415983, n_hidden=3, n_neurons=24 ...\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 70us/sample - loss: 2.9070 - val_loss: 5.0329\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 1.2882 - val_loss: 3.1420\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.9742 - val_loss: 1.9384\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.8517 - val_loss: 1.3691\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7864 - val_loss: 0.9828\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7436 - val_loss: 0.7970\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7132 - val_loss: 0.7109\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6891 - val_loss: 0.6661\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6690 - val_loss: 0.6342\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6512 - val_loss: 0.6156\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6353 - val_loss: 0.5987\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6208 - val_loss: 0.5851\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6073 - val_loss: 0.5730\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5951 - val_loss: 0.5607\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.5836 - val_loss: 0.5494\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5731 - val_loss: 0.5391\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.5633 - val_loss: 0.5295\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5540 - val_loss: 0.5217\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.5454 - val_loss: 0.5125\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5374 - val_loss: 0.5055\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5299 - val_loss: 0.4995\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5229 - val_loss: 0.4932\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5162 - val_loss: 0.4861\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5101 - val_loss: 0.4830\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5044 - val_loss: 0.4772\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4990 - val_loss: 0.4746\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 53us/sample - loss: 0.4940 - val_loss: 0.4719\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 49us/sample - loss: 0.4892 - val_loss: 0.4687\n",
      "Epoch 29/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.4848 - val_loss: 0.4641\n",
      "Epoch 30/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4805 - val_loss: 0.4613\n",
      "Epoch 31/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4766 - val_loss: 0.4599\n",
      "Epoch 32/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4729 - val_loss: 0.4561\n",
      "Epoch 33/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4694 - val_loss: 0.4555\n",
      "Epoch 34/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4660 - val_loss: 0.4585\n",
      "Epoch 35/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.4630 - val_loss: 0.4545\n",
      "Epoch 36/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4600 - val_loss: 0.4519\n",
      "Epoch 37/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4573 - val_loss: 0.4524\n",
      "Epoch 38/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4547 - val_loss: 0.4560\n",
      "Epoch 39/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4521 - val_loss: 0.4619\n",
      "Epoch 40/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4500 - val_loss: 0.4538\n",
      "Epoch 41/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4478 - val_loss: 0.4522\n",
      "Epoch 42/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4456 - val_loss: 0.4540\n",
      "Epoch 43/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4435 - val_loss: 0.4482\n",
      "Epoch 44/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4416 - val_loss: 0.4525\n",
      "Epoch 45/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4398 - val_loss: 0.4494\n",
      "Epoch 46/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4381 - val_loss: 0.4478\n",
      "Epoch 47/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4364 - val_loss: 0.4484\n",
      "Epoch 48/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4349 - val_loss: 0.4514\n",
      "Epoch 49/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4334 - val_loss: 0.4516\n",
      "Epoch 50/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4319 - val_loss: 0.4455\n",
      "Epoch 51/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4304 - val_loss: 0.4405\n",
      "Epoch 52/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4292 - val_loss: 0.4437\n",
      "Epoch 53/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4277 - val_loss: 0.4417\n",
      "Epoch 54/100\n",
      "7740/7740 [==============================] - 0s 52us/sample - loss: 0.4266 - val_loss: 0.4433\n",
      "Epoch 55/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4252 - val_loss: 0.4468\n",
      "Epoch 56/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4241 - val_loss: 0.4536\n",
      "Epoch 57/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4228 - val_loss: 0.4470\n",
      "Epoch 58/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4216 - val_loss: 0.4569\n",
      "Epoch 59/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4207 - val_loss: 0.4554\n",
      "Epoch 60/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4197 - val_loss: 0.4547\n",
      "Epoch 61/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4187 - val_loss: 0.4475\n",
      "3870/3870 [==============================] - 0s 20us/sample - loss: 0.4175\n",
      "[CV]  learning_rate=0.0003920021771415983, n_hidden=3, n_neurons=24, total=  21.4s\n",
      "[CV] learning_rate=0.006010328378268217, n_hidden=0, n_neurons=2 .....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 59us/sample - loss: 2.1211 - val_loss: 21.1308\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.9808 - val_loss: 95.9915\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 1.6776 - val_loss: 221.0205\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 3.8367 - val_loss: 450.6542\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 49us/sample - loss: 3.9032 - val_loss: 933.3998\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 51us/sample - loss: 11.5679 - val_loss: 2256.4544\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 37.9562 - val_loss: 5370.5414\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 37.4190 - val_loss: 13317.2301\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 73.9304 - val_loss: 29245.5294\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 213.6844 - val_loss: 67753.4594\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 462.1570 - val_loss: 152754.9162\n",
      "3870/3870 [==============================] - 0s 20us/sample - loss: 406.6080\n",
      "[CV]  learning_rate=0.006010328378268217, n_hidden=0, n_neurons=2, total=   3.9s\n",
      "[CV] learning_rate=0.006010328378268217, n_hidden=0, n_neurons=2 .....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 62us/sample - loss: 1.4677 - val_loss: 13.3385\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5767 - val_loss: 16.4312\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5515 - val_loss: 18.4041\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5368 - val_loss: 18.4688\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.5278 - val_loss: 18.5664\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5201 - val_loss: 18.5914\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5141 - val_loss: 18.0228\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5116 - val_loss: 18.2423\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.5096 - val_loss: 18.0343\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5066 - val_loss: 18.6712\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.5062 - val_loss: 17.8861\n",
      "3870/3870 [==============================] - 0s 21us/sample - loss: 0.9307\n",
      "[CV]  learning_rate=0.006010328378268217, n_hidden=0, n_neurons=2, total=   3.8s\n",
      "[CV] learning_rate=0.006010328378268217, n_hidden=0, n_neurons=2 .....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 83us/sample - loss: 1.8572 - val_loss: 0.9861\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5987 - val_loss: 0.9252\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 37us/sample - loss: 0.5559 - val_loss: 3.3524\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.6727 - val_loss: 9.0939\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.6038 - val_loss: 19.4521\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.8763 - val_loss: 28.0253\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.8241 - val_loss: 45.1730\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 1.4897 - val_loss: 47.6945\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 37us/sample - loss: 1.3526 - val_loss: 94.9098\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 2.5568 - val_loss: 105.7355\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 1.3688 - val_loss: 134.7788\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 2.9960 - val_loss: 193.8962\n",
      "3870/3870 [==============================] - 0s 18us/sample - loss: 0.9807\n",
      "[CV]  learning_rate=0.006010328378268217, n_hidden=0, n_neurons=2, total=   4.3s\n",
      "[CV] learning_rate=0.008339092654580042, n_hidden=1, n_neurons=38 ....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 61us/sample - loss: 1.2569 - val_loss: 40.1895\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.9780 - val_loss: 196.2189\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 1.7930 - val_loss: 2.4461\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.4490 - val_loss: 1.9442\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4378 - val_loss: 14.0566\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5361 - val_loss: 25.7459\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.6511 - val_loss: 0.3580\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3706 - val_loss: 0.3511\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3659 - val_loss: 0.3543\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3618 - val_loss: 0.3492\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.3581 - val_loss: 0.3460\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.3550 - val_loss: 0.3574\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.3531 - val_loss: 0.3557\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.3503 - val_loss: 0.4685\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3494 - val_loss: 0.3399\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3485 - val_loss: 0.3644\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3462 - val_loss: 0.3809\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3451 - val_loss: 0.3840\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3438 - val_loss: 0.3788\n",
      "Epoch 20/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3429 - val_loss: 0.3633\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3408 - val_loss: 0.3366\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.3400 - val_loss: 0.3583\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3397 - val_loss: 0.3759\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3373 - val_loss: 0.3592\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3364 - val_loss: 0.3754\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3354 - val_loss: 0.3558\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3343 - val_loss: 0.3720\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3324 - val_loss: 0.3559\n",
      "Epoch 29/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3298 - val_loss: 0.3524\n",
      "Epoch 30/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3285 - val_loss: 0.3484\n",
      "Epoch 31/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3260 - val_loss: 0.3706\n",
      "3870/3870 [==============================] - 0s 22us/sample - loss: 0.3526\n",
      "[CV]  learning_rate=0.008339092654580042, n_hidden=1, n_neurons=38, total=  10.1s\n",
      "[CV] learning_rate=0.008339092654580042, n_hidden=1, n_neurons=38 ....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 63us/sample - loss: 0.8947 - val_loss: 3.5465\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5222 - val_loss: 0.4637\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4628 - val_loss: 0.4256\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4362 - val_loss: 0.4192\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4215 - val_loss: 0.4591\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4122 - val_loss: 0.5954\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.4047 - val_loss: 0.6497\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4000 - val_loss: 0.7834\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3957 - val_loss: 0.8585\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3918 - val_loss: 1.0683\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3882 - val_loss: 0.9851\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3836 - val_loss: 0.9689\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3830 - val_loss: 1.1161\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3792 - val_loss: 1.3615\n",
      "3870/3870 [==============================] - 0s 22us/sample - loss: 0.4046\n",
      "[CV]  learning_rate=0.008339092654580042, n_hidden=1, n_neurons=38, total=   4.8s\n",
      "[CV] learning_rate=0.008339092654580042, n_hidden=1, n_neurons=38 ....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 61us/sample - loss: 1.1191 - val_loss: 13.5203\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5587 - val_loss: 45.9143\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 1.1573 - val_loss: 11.7740\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.6116 - val_loss: 13.3710\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5644 - val_loss: 0.3805\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4145 - val_loss: 0.3733\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4124 - val_loss: 0.3731\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4161 - val_loss: 0.3624\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3921 - val_loss: 0.3585\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3875 - val_loss: 0.3541\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3847 - val_loss: 0.3531\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3832 - val_loss: 0.3501\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3766 - val_loss: 0.3481\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3746 - val_loss: 0.3682\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3772 - val_loss: 0.3692\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3721 - val_loss: 0.3517\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3678 - val_loss: 0.3814\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3773 - val_loss: 0.3455\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3655 - val_loss: 0.3699\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3712 - val_loss: 0.3731\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3626 - val_loss: 0.3615\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3615 - val_loss: 0.3422\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3610 - val_loss: 0.3518\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3582 - val_loss: 0.3447\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3570 - val_loss: 0.3711\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3559 - val_loss: 0.3622\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3563 - val_loss: 0.3662\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3863 - val_loss: 0.3685\n",
      "Epoch 29/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3762 - val_loss: 0.3449\n",
      "Epoch 30/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3641 - val_loss: 0.3400\n",
      "Epoch 31/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3547 - val_loss: 0.3316\n",
      "Epoch 32/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3587 - val_loss: 0.3810\n",
      "Epoch 33/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3508 - val_loss: 0.3341\n",
      "Epoch 34/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3500 - val_loss: 0.3811\n",
      "Epoch 35/100\n",
      "7740/7740 [==============================] - 0s 49us/sample - loss: 0.3485 - val_loss: 0.3379\n",
      "Epoch 36/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3503 - val_loss: 0.3418\n",
      "Epoch 37/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.3468 - val_loss: 0.3626\n",
      "Epoch 38/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3515 - val_loss: 0.3792\n",
      "Epoch 39/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3478 - val_loss: 0.3299\n",
      "Epoch 40/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3450 - val_loss: 0.3286\n",
      "Epoch 41/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3465 - val_loss: 0.3359\n",
      "Epoch 42/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3423 - val_loss: 0.3507\n",
      "Epoch 43/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3542 - val_loss: 0.3315\n",
      "Epoch 44/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3543 - val_loss: 0.3362\n",
      "Epoch 45/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3433 - val_loss: 0.3447\n",
      "Epoch 46/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3438 - val_loss: 0.3555\n",
      "Epoch 47/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3457 - val_loss: 0.3562\n",
      "Epoch 48/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3399 - val_loss: 0.3606\n",
      "Epoch 49/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3391 - val_loss: 0.3702\n",
      "Epoch 50/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4423 - val_loss: 0.3543\n",
      "3870/3870 [==============================] - 0s 20us/sample - loss: 0.3416\n",
      "[CV]  learning_rate=0.008339092654580042, n_hidden=1, n_neurons=38, total=  16.3s\n",
      "[CV] learning_rate=0.00030107783636342726, n_hidden=3, n_neurons=21 ..\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 69us/sample - loss: 4.0249 - val_loss: 8.1049\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 2.3117 - val_loss: 7.7845\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 1.6343 - val_loss: 5.9924\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 1.3424 - val_loss: 4.1657\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 1.1817 - val_loss: 2.7137\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 1.0700 - val_loss: 2.1602\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.9890 - val_loss: 1.7833\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.9245 - val_loss: 1.4592\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.8705 - val_loss: 1.2284\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.8264 - val_loss: 1.0696\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7890 - val_loss: 0.9366\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7581 - val_loss: 0.8372\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.7322 - val_loss: 0.7673\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7103 - val_loss: 0.7110\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6923 - val_loss: 0.6765\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6765 - val_loss: 0.6552\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6633 - val_loss: 0.6374\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6518 - val_loss: 0.6275\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.6418 - val_loss: 0.6202\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6327 - val_loss: 0.6147\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6245 - val_loss: 0.6085\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6168 - val_loss: 0.6056\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6096 - val_loss: 0.6016\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6028 - val_loss: 0.5962\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5963 - val_loss: 0.5940\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5902 - val_loss: 0.5894\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5840 - val_loss: 0.5902\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5785 - val_loss: 0.5810\n",
      "Epoch 29/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5729 - val_loss: 0.5763\n",
      "Epoch 30/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5675 - val_loss: 0.5734\n",
      "Epoch 31/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.5622 - val_loss: 0.5674\n",
      "Epoch 32/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.5570 - val_loss: 0.5654\n",
      "Epoch 33/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5520 - val_loss: 0.5583\n",
      "Epoch 34/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5470 - val_loss: 0.5519\n",
      "Epoch 35/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5422 - val_loss: 0.5506\n",
      "Epoch 36/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5373 - val_loss: 0.5506\n",
      "Epoch 37/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5332 - val_loss: 0.5409\n",
      "Epoch 38/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5285 - val_loss: 0.5344\n",
      "Epoch 39/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5242 - val_loss: 0.5294\n",
      "Epoch 40/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5199 - val_loss: 0.5237\n",
      "Epoch 41/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5158 - val_loss: 0.5209\n",
      "Epoch 42/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5117 - val_loss: 0.5159\n",
      "Epoch 43/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5076 - val_loss: 0.5117\n",
      "Epoch 44/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5038 - val_loss: 0.5077\n",
      "Epoch 45/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5000 - val_loss: 0.5019\n",
      "Epoch 46/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4964 - val_loss: 0.4978\n",
      "Epoch 47/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4927 - val_loss: 0.4916\n",
      "Epoch 48/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4890 - val_loss: 0.4886\n",
      "Epoch 49/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4855 - val_loss: 0.4859\n",
      "Epoch 50/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4823 - val_loss: 0.4793\n",
      "Epoch 51/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.4788 - val_loss: 0.4758\n",
      "Epoch 52/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4755 - val_loss: 0.4715\n",
      "Epoch 53/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4723 - val_loss: 0.4689\n",
      "Epoch 54/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4692 - val_loss: 0.4640\n",
      "Epoch 55/100\n",
      "7740/7740 [==============================] - 0s 49us/sample - loss: 0.4660 - val_loss: 0.4613\n",
      "Epoch 56/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4631 - val_loss: 0.4573\n",
      "Epoch 57/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4601 - val_loss: 0.4542\n",
      "Epoch 58/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4571 - val_loss: 0.4492\n",
      "Epoch 59/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4544 - val_loss: 0.4461\n",
      "Epoch 60/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4517 - val_loss: 0.4431\n",
      "Epoch 61/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4489 - val_loss: 0.4396\n",
      "Epoch 62/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4462 - val_loss: 0.4378\n",
      "Epoch 63/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.4440 - val_loss: 0.4337\n",
      "Epoch 64/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4415 - val_loss: 0.4308\n",
      "Epoch 65/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4390 - val_loss: 0.4297\n",
      "Epoch 66/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4369 - val_loss: 0.4260\n",
      "Epoch 67/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4346 - val_loss: 0.4239\n",
      "Epoch 68/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4324 - val_loss: 0.4212\n",
      "Epoch 69/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4302 - val_loss: 0.4192\n",
      "Epoch 70/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4282 - val_loss: 0.4172\n",
      "Epoch 71/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4260 - val_loss: 0.4151\n",
      "Epoch 72/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4243 - val_loss: 0.4132\n",
      "Epoch 73/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4224 - val_loss: 0.4112\n",
      "Epoch 74/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4204 - val_loss: 0.4092\n",
      "Epoch 75/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4188 - val_loss: 0.4074\n",
      "Epoch 76/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4170 - val_loss: 0.4056\n",
      "Epoch 77/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4152 - val_loss: 0.4039\n",
      "Epoch 78/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.4135 - val_loss: 0.4026\n",
      "Epoch 79/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.4119 - val_loss: 0.4013\n",
      "Epoch 80/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4105 - val_loss: 0.3995\n",
      "Epoch 81/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4090 - val_loss: 0.3980\n",
      "Epoch 82/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4075 - val_loss: 0.3969\n",
      "Epoch 83/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4060 - val_loss: 0.3963\n",
      "Epoch 84/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4046 - val_loss: 0.3949\n",
      "Epoch 85/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4033 - val_loss: 0.3935\n",
      "Epoch 86/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4020 - val_loss: 0.3932\n",
      "Epoch 87/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4005 - val_loss: 0.3933\n",
      "Epoch 88/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3993 - val_loss: 0.3905\n",
      "Epoch 89/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3982 - val_loss: 0.3899\n",
      "Epoch 90/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3970 - val_loss: 0.3893\n",
      "Epoch 91/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3958 - val_loss: 0.3888\n",
      "Epoch 92/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3947 - val_loss: 0.3887\n",
      "Epoch 93/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3936 - val_loss: 0.3880\n",
      "Epoch 94/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3926 - val_loss: 0.3868\n",
      "Epoch 95/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.3914 - val_loss: 0.3851\n",
      "Epoch 96/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3906 - val_loss: 0.3863\n",
      "Epoch 97/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3896 - val_loss: 0.3865\n",
      "Epoch 98/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.3887 - val_loss: 0.3859\n",
      "Epoch 99/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3877 - val_loss: 0.3848\n",
      "Epoch 100/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3870 - val_loss: 0.3835\n",
      "3870/3870 [==============================] - 0s 21us/sample - loss: 0.3991\n",
      "[CV]  learning_rate=0.00030107783636342726, n_hidden=3, n_neurons=21, total=  34.5s\n",
      "[CV] learning_rate=0.00030107783636342726, n_hidden=3, n_neurons=21 ..\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 70us/sample - loss: 5.0596 - val_loss: 2.9550\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 2.1383 - val_loss: 5.8779\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 1.2774 - val_loss: 10.6370\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 1.0897 - val_loss: 11.0355\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 1.0049 - val_loss: 9.7557\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.9441 - val_loss: 8.1254\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.8975 - val_loss: 6.3476\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.8607 - val_loss: 4.8009\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.8306 - val_loss: 3.4145\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.8060 - val_loss: 2.6140\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.7845 - val_loss: 1.8966\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7657 - val_loss: 1.3967\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7491 - val_loss: 1.1024\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7339 - val_loss: 0.8552\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7199 - val_loss: 0.7101\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.7073 - val_loss: 0.6640\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6950 - val_loss: 0.6902\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6835 - val_loss: 0.7543\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6729 - val_loss: 0.8957\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6624 - val_loss: 1.1091\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6528 - val_loss: 1.3141\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.6435 - val_loss: 1.5836\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6348 - val_loss: 1.8804\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 49us/sample - loss: 0.6266 - val_loss: 2.1592\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.6188 - val_loss: 2.5489\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6112 - val_loss: 2.8359\n",
      "3870/3870 [==============================] - 0s 20us/sample - loss: 0.6766\n",
      "[CV]  learning_rate=0.00030107783636342726, n_hidden=3, n_neurons=21, total=   9.2s\n",
      "[CV] learning_rate=0.00030107783636342726, n_hidden=3, n_neurons=21 ..\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 70us/sample - loss: 4.4070 - val_loss: 3.5477\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 2.5644 - val_loss: 3.0315\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 1.4006 - val_loss: 2.9878\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.9858 - val_loss: 1.9229\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.8391 - val_loss: 1.3465\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.7683 - val_loss: 1.0212\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7267 - val_loss: 0.8280\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6991 - val_loss: 0.7424\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6798 - val_loss: 0.6976\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6650 - val_loss: 0.6660\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6528 - val_loss: 0.6406\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6422 - val_loss: 0.6224\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6325 - val_loss: 0.6083\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6236 - val_loss: 0.5963\n",
      "Epoch 15/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6152 - val_loss: 0.5853\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.6072 - val_loss: 0.5758\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5993 - val_loss: 0.5672\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5916 - val_loss: 0.5599\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5842 - val_loss: 0.5519\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.5769 - val_loss: 0.5446\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5698 - val_loss: 0.5376\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5628 - val_loss: 0.5309\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5559 - val_loss: 0.5241\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5492 - val_loss: 0.5178\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5426 - val_loss: 0.5111\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5361 - val_loss: 0.5050\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5298 - val_loss: 0.4990\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.5236 - val_loss: 0.4929\n",
      "Epoch 29/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5175 - val_loss: 0.4875\n",
      "Epoch 30/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5116 - val_loss: 0.4819\n",
      "Epoch 31/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.5059 - val_loss: 0.4775\n",
      "Epoch 32/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5005 - val_loss: 0.4727\n",
      "Epoch 33/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4953 - val_loss: 0.4684\n",
      "Epoch 34/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4902 - val_loss: 0.4652\n",
      "Epoch 35/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4856 - val_loss: 0.4610\n",
      "Epoch 36/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4811 - val_loss: 0.4579\n",
      "Epoch 37/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4768 - val_loss: 0.4553\n",
      "Epoch 38/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4728 - val_loss: 0.4532\n",
      "Epoch 39/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4689 - val_loss: 0.4517\n",
      "Epoch 40/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4653 - val_loss: 0.4492\n",
      "Epoch 41/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4618 - val_loss: 0.4478\n",
      "Epoch 42/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4586 - val_loss: 0.4464\n",
      "Epoch 43/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4554 - val_loss: 0.4451\n",
      "Epoch 44/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.4525 - val_loss: 0.4445\n",
      "Epoch 45/100\n",
      "7740/7740 [==============================] - 0s 48us/sample - loss: 0.4496 - val_loss: 0.4439\n",
      "Epoch 46/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4470 - val_loss: 0.4439\n",
      "Epoch 47/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4445 - val_loss: 0.4430\n",
      "Epoch 48/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4421 - val_loss: 0.4435\n",
      "Epoch 49/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4398 - val_loss: 0.4437\n",
      "Epoch 50/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4376 - val_loss: 0.4440\n",
      "Epoch 51/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4354 - val_loss: 0.4437\n",
      "Epoch 52/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4335 - val_loss: 0.4447\n",
      "Epoch 53/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4315 - val_loss: 0.4446\n",
      "Epoch 54/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4297 - val_loss: 0.4453\n",
      "Epoch 55/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4280 - val_loss: 0.4471\n",
      "Epoch 56/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4263 - val_loss: 0.4492\n",
      "Epoch 57/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4246 - val_loss: 0.4486\n",
      "3870/3870 [==============================] - 0s 21us/sample - loss: 0.4258\n",
      "[CV]  learning_rate=0.00030107783636342726, n_hidden=3, n_neurons=21, total=  19.5s\n",
      "[CV] learning_rate=0.005153286333701512, n_hidden=1, n_neurons=22 ....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 62us/sample - loss: 1.4845 - val_loss: 35.5339\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.8461 - val_loss: 0.5467\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5170 - val_loss: 0.4691\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4755 - val_loss: 0.4427\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4501 - val_loss: 0.4391\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4353 - val_loss: 0.4316\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4246 - val_loss: 0.4399\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4170 - val_loss: 0.4396\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4120 - val_loss: 0.4404\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4074 - val_loss: 0.4411\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4038 - val_loss: 0.4408\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4013 - val_loss: 0.4403\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3988 - val_loss: 0.4407\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3962 - val_loss: 0.4424\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3945 - val_loss: 0.4410\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3931 - val_loss: 0.4383\n",
      "3870/3870 [==============================] - 0s 20us/sample - loss: 0.3994\n",
      "[CV]  learning_rate=0.005153286333701512, n_hidden=1, n_neurons=22, total=   5.3s\n",
      "[CV] learning_rate=0.005153286333701512, n_hidden=1, n_neurons=22 ....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 63us/sample - loss: 1.2414 - val_loss: 0.6622\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5775 - val_loss: 0.9135\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.5142 - val_loss: 1.3277\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4808 - val_loss: 1.0918\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4607 - val_loss: 0.9326\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4453 - val_loss: 0.6336\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4315 - val_loss: 0.5175\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4235 - val_loss: 0.4141\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4159 - val_loss: 0.3806\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4089 - val_loss: 0.3892\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4051 - val_loss: 0.4235\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3998 - val_loss: 0.4262\n",
      "Epoch 13/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3988 - val_loss: 0.4753\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3943 - val_loss: 0.5508\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3902 - val_loss: 0.5443\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3896 - val_loss: 0.6561\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3867 - val_loss: 0.6874\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3837 - val_loss: 0.7188\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3818 - val_loss: 0.6562\n",
      "3870/3870 [==============================] - 0s 22us/sample - loss: 0.3968\n",
      "[CV]  learning_rate=0.005153286333701512, n_hidden=1, n_neurons=22, total=   6.3s\n",
      "[CV] learning_rate=0.005153286333701512, n_hidden=1, n_neurons=22 ....\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 64us/sample - loss: 1.6836 - val_loss: 51.2996\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.9003 - val_loss: 0.5389\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5341 - val_loss: 0.4896\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4987 - val_loss: 0.4567\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4742 - val_loss: 0.4354\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4585 - val_loss: 0.4215\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4479 - val_loss: 0.4123\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4401 - val_loss: 0.4088\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4336 - val_loss: 0.4047\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4284 - val_loss: 0.4023\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4241 - val_loss: 0.4012\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4208 - val_loss: 0.4043\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4175 - val_loss: 0.4022\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4147 - val_loss: 0.4040\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4124 - val_loss: 0.4049\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4098 - val_loss: 0.4034\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4073 - val_loss: 0.4036\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4050 - val_loss: 0.4038\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4031 - val_loss: 0.4000\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.4014 - val_loss: 0.4091\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3995 - val_loss: 0.4060\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3979 - val_loss: 0.4134\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3958 - val_loss: 0.4069\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3941 - val_loss: 0.4115\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3928 - val_loss: 0.4106\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3914 - val_loss: 0.4099\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3899 - val_loss: 0.4136\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3890 - val_loss: 0.4096\n",
      "Epoch 29/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.3879 - val_loss: 0.4087\n",
      "3870/3870 [==============================] - 0s 20us/sample - loss: 0.3854\n",
      "[CV]  learning_rate=0.005153286333701512, n_hidden=1, n_neurons=22, total=   9.5s\n",
      "[CV] learning_rate=0.0003099230412972121, n_hidden=0, n_neurons=49 ...\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 58us/sample - loss: 7.7217 - val_loss: 43.0506\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 5.5388 - val_loss: 27.4292\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 4.0672 - val_loss: 17.4115\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 3.0653 - val_loss: 11.0238\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 2.3732 - val_loss: 7.0240\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 1.8924 - val_loss: 4.5011\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 1.5535 - val_loss: 2.9137\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 1.3132 - val_loss: 1.9725\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 1.1424 - val_loss: 1.4086\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 1.0195 - val_loss: 1.0738\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.9299 - val_loss: 0.8882\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.8642 - val_loss: 0.7914\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.8156 - val_loss: 0.7516\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.7794 - val_loss: 0.7475\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.7520 - val_loss: 0.7645\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 37us/sample - loss: 0.7311 - val_loss: 0.7813\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.7149 - val_loss: 0.8135\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.7023 - val_loss: 0.8417\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.6923 - val_loss: 0.8599\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.6841 - val_loss: 0.8713\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.6773 - val_loss: 0.8857\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.6715 - val_loss: 0.9026\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.6664 - val_loss: 0.9227\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6621 - val_loss: 0.9242\n",
      "3870/3870 [==============================] - 0s 19us/sample - loss: 0.6633\n",
      "[CV]  learning_rate=0.0003099230412972121, n_hidden=0, n_neurons=49, total=   7.6s\n",
      "[CV] learning_rate=0.0003099230412972121, n_hidden=0, n_neurons=49 ...\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 58us/sample - loss: 7.6319 - val_loss: 25.5299\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 5.6945 - val_loss: 23.7997\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 4.3241 - val_loss: 22.5858\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 3.3431 - val_loss: 21.7425\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 2.6344 - val_loss: 21.1466\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 2.1176 - val_loss: 20.7281\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 1.7381 - val_loss: 20.4354\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 1.4574 - val_loss: 20.2416\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 1.2484 - val_loss: 20.1131\n",
      "Epoch 10/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 39us/sample - loss: 1.0921 - val_loss: 20.0378\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.9744 - val_loss: 19.9971\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.8855 - val_loss: 19.9839\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.8178 - val_loss: 19.9964\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.7659 - val_loss: 20.0215\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.7260 - val_loss: 20.0579\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.6950 - val_loss: 20.1105\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.6709 - val_loss: 20.1648\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6518 - val_loss: 20.2247\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6366 - val_loss: 20.2845\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.6245 - val_loss: 20.3550\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6147 - val_loss: 20.4260\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6067 - val_loss: 20.4958\n",
      "3870/3870 [==============================] - 0s 19us/sample - loss: 1.0972\n",
      "[CV]  learning_rate=0.0003099230412972121, n_hidden=0, n_neurons=49, total=   7.0s\n",
      "[CV] learning_rate=0.0003099230412972121, n_hidden=0, n_neurons=49 ...\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 0s 58us/sample - loss: 6.1534 - val_loss: 7.6082\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 4.4875 - val_loss: 4.9014\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 3.3698 - val_loss: 3.2853\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 2.6022 - val_loss: 2.3426\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 2.0664 - val_loss: 1.7733\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 1.6865 - val_loss: 1.4371\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 1.4144 - val_loss: 1.2377\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 1.2173 - val_loss: 1.1218\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 37us/sample - loss: 1.0732 - val_loss: 1.0580\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.9674 - val_loss: 1.0234\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.8892 - val_loss: 0.9967\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.8310 - val_loss: 0.9828\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.7874 - val_loss: 0.9636\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.7544 - val_loss: 0.9520\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.7294 - val_loss: 0.9401\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 37us/sample - loss: 0.7101 - val_loss: 0.9402\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.6951 - val_loss: 0.9493\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6833 - val_loss: 0.9449\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6738 - val_loss: 0.9472\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6660 - val_loss: 0.9560\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6597 - val_loss: 0.9580\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 40us/sample - loss: 0.6543 - val_loss: 0.9593\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6496 - val_loss: 0.9608\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 38us/sample - loss: 0.6456 - val_loss: 0.9579\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 39us/sample - loss: 0.6419 - val_loss: 0.9564\n",
      "3870/3870 [==============================] - 0s 20us/sample - loss: 0.6453\n",
      "[CV]  learning_rate=0.0003099230412972121, n_hidden=0, n_neurons=49, total=   7.9s\n",
      "[CV] learning_rate=0.0033625641252688094, n_hidden=2, n_neurons=42 ...\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 69us/sample - loss: 1.3940 - val_loss: 12.7105\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.7302 - val_loss: 4.0264\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.5546 - val_loss: 0.6581\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4725 - val_loss: 0.4863\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4407 - val_loss: 0.4827\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4230 - val_loss: 0.4355\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4089 - val_loss: 0.4474\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4005 - val_loss: 0.4165\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3942 - val_loss: 0.4195\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3864 - val_loss: 0.4090\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3800 - val_loss: 0.3859\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3767 - val_loss: 0.3827\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3722 - val_loss: 0.3834\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3676 - val_loss: 0.3921\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3643 - val_loss: 0.3808\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3632 - val_loss: 0.3849\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3595 - val_loss: 0.3894\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3570 - val_loss: 0.3844\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3560 - val_loss: 0.3801\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3534 - val_loss: 0.3620\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3507 - val_loss: 0.3653\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3495 - val_loss: 0.3751\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3480 - val_loss: 0.3661\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3457 - val_loss: 0.3620\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3439 - val_loss: 0.3773\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3441 - val_loss: 0.3463\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3432 - val_loss: 0.3723\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3405 - val_loss: 0.3710\n",
      "Epoch 29/100\n",
      "7740/7740 [==============================] - 0s 46us/sample - loss: 0.3384 - val_loss: 0.3462\n",
      "Epoch 30/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3375 - val_loss: 0.3574\n",
      "Epoch 31/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3359 - val_loss: 0.3712\n",
      "Epoch 32/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3349 - val_loss: 0.3750\n",
      "Epoch 33/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3345 - val_loss: 0.3333\n",
      "Epoch 34/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3332 - val_loss: 0.3516\n",
      "Epoch 35/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3315 - val_loss: 0.3521\n",
      "Epoch 36/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3310 - val_loss: 0.3373\n",
      "Epoch 37/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3300 - val_loss: 0.3442\n",
      "Epoch 38/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3284 - val_loss: 0.3440\n",
      "Epoch 39/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3280 - val_loss: 0.3466\n",
      "Epoch 40/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3275 - val_loss: 0.3503\n",
      "Epoch 41/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3269 - val_loss: 0.3352\n",
      "Epoch 42/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3261 - val_loss: 0.3405\n",
      "Epoch 43/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3250 - val_loss: 0.3611\n",
      "3870/3870 [==============================] - 0s 20us/sample - loss: 0.3648\n",
      "[CV]  learning_rate=0.0033625641252688094, n_hidden=2, n_neurons=42, total=  14.5s\n",
      "[CV] learning_rate=0.0033625641252688094, n_hidden=2, n_neurons=42 ...\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 67us/sample - loss: 1.1931 - val_loss: 0.9394\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.6049 - val_loss: 0.8525\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.5370 - val_loss: 1.1657\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.4909 - val_loss: 0.9411\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 45us/sample - loss: 0.4622 - val_loss: 0.7834\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.4407 - val_loss: 0.5922\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4252 - val_loss: 0.4672\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4137 - val_loss: 0.3866\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4047 - val_loss: 0.3882\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3967 - val_loss: 0.4710\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3903 - val_loss: 0.4998\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3837 - val_loss: 0.6031\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3797 - val_loss: 0.6998\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3754 - val_loss: 0.8516\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3709 - val_loss: 0.7903\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3683 - val_loss: 0.9818\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3648 - val_loss: 1.0229\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3628 - val_loss: 1.0536\n",
      "3870/3870 [==============================] - 0s 21us/sample - loss: 0.3868\n",
      "[CV]  learning_rate=0.0033625641252688094, n_hidden=2, n_neurons=42, total=   6.3s\n",
      "[CV] learning_rate=0.0033625641252688094, n_hidden=2, n_neurons=42 ...\n",
      "Train on 7740 samples, validate on 3870 samples\n",
      "Epoch 1/100\n",
      "7740/7740 [==============================] - 1s 65us/sample - loss: 1.1176 - val_loss: 3.6631\n",
      "Epoch 2/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.7068 - val_loss: 2.4415\n",
      "Epoch 3/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5988 - val_loss: 0.7542\n",
      "Epoch 4/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.5253 - val_loss: 0.8354\n",
      "Epoch 5/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4824 - val_loss: 0.4450\n",
      "Epoch 6/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4547 - val_loss: 0.4508\n",
      "Epoch 7/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4355 - val_loss: 0.4923\n",
      "Epoch 8/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.4245 - val_loss: 0.4276\n",
      "Epoch 9/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.4133 - val_loss: 0.5228\n",
      "Epoch 10/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.4061 - val_loss: 0.4544\n",
      "Epoch 11/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3993 - val_loss: 0.3845\n",
      "Epoch 12/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3928 - val_loss: 0.5825\n",
      "Epoch 13/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3898 - val_loss: 0.4636\n",
      "Epoch 14/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3842 - val_loss: 0.6485\n",
      "Epoch 15/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3808 - val_loss: 0.5205\n",
      "Epoch 16/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3764 - val_loss: 0.5565\n",
      "Epoch 17/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3740 - val_loss: 0.3703\n",
      "Epoch 18/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3700 - val_loss: 0.3460\n",
      "Epoch 19/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3663 - val_loss: 0.6020\n",
      "Epoch 20/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3701 - val_loss: 0.5423\n",
      "Epoch 21/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3631 - val_loss: 0.3545\n",
      "Epoch 22/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3601 - val_loss: 0.3745\n",
      "Epoch 23/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3576 - val_loss: 0.3672\n",
      "Epoch 24/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3553 - val_loss: 0.3775\n",
      "Epoch 25/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3531 - val_loss: 0.5842\n",
      "Epoch 26/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3537 - val_loss: 0.3427\n",
      "Epoch 27/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3499 - val_loss: 0.4517\n",
      "Epoch 28/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3520 - val_loss: 0.3971\n",
      "Epoch 29/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3483 - val_loss: 0.3307\n",
      "Epoch 30/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3481 - val_loss: 0.3352\n",
      "Epoch 31/100\n",
      "7740/7740 [==============================] - 0s 43us/sample - loss: 0.3458 - val_loss: 0.3398\n",
      "Epoch 32/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3438 - val_loss: 0.5529\n",
      "Epoch 33/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.3440 - val_loss: 0.3624\n",
      "Epoch 34/100\n",
      "7740/7740 [==============================] - 0s 47us/sample - loss: 0.3413 - val_loss: 0.5745\n",
      "Epoch 35/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3429 - val_loss: 0.3335\n",
      "Epoch 36/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3395 - val_loss: 0.3256\n",
      "Epoch 37/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3388 - val_loss: 0.6801\n",
      "Epoch 38/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3424 - val_loss: 0.3271\n",
      "Epoch 39/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3362 - val_loss: 0.8728\n",
      "Epoch 40/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3417 - val_loss: 1.1699\n",
      "Epoch 41/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3532 - val_loss: 1.4036\n",
      "Epoch 42/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3477 - val_loss: 0.4005\n",
      "Epoch 43/100\n",
      "7740/7740 [==============================] - 0s 44us/sample - loss: 0.3387 - val_loss: 0.6377\n",
      "Epoch 44/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3354 - val_loss: 0.3228\n",
      "Epoch 45/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3324 - val_loss: 0.4816\n",
      "Epoch 46/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3331 - val_loss: 0.4228\n",
      "Epoch 47/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3331 - val_loss: 0.6872\n",
      "Epoch 48/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3319 - val_loss: 0.3177\n",
      "Epoch 49/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3295 - val_loss: 0.6891\n",
      "Epoch 50/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3376 - val_loss: 0.5869\n",
      "Epoch 51/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3310 - val_loss: 0.5921\n",
      "Epoch 52/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3296 - val_loss: 0.4285\n",
      "Epoch 53/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3284 - val_loss: 0.4579\n",
      "Epoch 54/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3287 - val_loss: 0.3434\n",
      "Epoch 55/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3288 - val_loss: 0.4285\n",
      "Epoch 56/100\n",
      "7740/7740 [==============================] - 0s 42us/sample - loss: 0.3257 - val_loss: 0.4101\n",
      "Epoch 57/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3243 - val_loss: 0.3204\n",
      "Epoch 58/100\n",
      "7740/7740 [==============================] - 0s 41us/sample - loss: 0.3236 - val_loss: 0.4966\n",
      "3870/3870 [==============================] - 0s 20us/sample - loss: 0.3270\n",
      "[CV]  learning_rate=0.0033625641252688094, n_hidden=2, n_neurons=42, total=  19.3s\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done  30 out of  30 | elapsed:  5.9min finished\n"
     ]
    },
    {
     "ename": "RuntimeError",
     "evalue": "Cannot clone object <tensorflow.python.keras.wrappers.scikit_learn.KerasRegressor object at 0x000001C61600AD48>, as the constructor either does not set or modifies parameter learning_rate",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mRuntimeError\u001b[0m                              Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-90-fe5bdfa90542>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m     11\u001b[0m rnd_search_cv.fit(X_train, y_train, epochs=100,\n\u001b[0;32m     12\u001b[0m                   \u001b[0mvalidation_data\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mX_valid\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0my_valid\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 13\u001b[1;33m                   callbacks=[keras.callbacks.EarlyStopping(patience=10)])\n\u001b[0m",
      "\u001b[1;32m~\\.conda\\envs\\atml_202021\\lib\\site-packages\\sklearn\\utils\\validation.py\u001b[0m in \u001b[0;36minner_f\u001b[1;34m(*args, **kwargs)\u001b[0m\n\u001b[0;32m     70\u001b[0m                           FutureWarning)\n\u001b[0;32m     71\u001b[0m         \u001b[0mkwargs\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mupdate\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m{\u001b[0m\u001b[0mk\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0marg\u001b[0m \u001b[1;32mfor\u001b[0m \u001b[0mk\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0marg\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mzip\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msig\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mparameters\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0margs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m}\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 72\u001b[1;33m         \u001b[1;32mreturn\u001b[0m \u001b[0mf\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     73\u001b[0m     \u001b[1;32mreturn\u001b[0m \u001b[0minner_f\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     74\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\.conda\\envs\\atml_202021\\lib\\site-packages\\sklearn\\model_selection\\_search.py\u001b[0m in \u001b[0;36mfit\u001b[1;34m(self, X, y, groups, **fit_params)\u001b[0m\n\u001b[0;32m    760\u001b[0m             \u001b[1;31m# of the params are estimators as well.\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    761\u001b[0m             self.best_estimator_ = clone(clone(base_estimator).set_params(\n\u001b[1;32m--> 762\u001b[1;33m                 **self.best_params_))\n\u001b[0m\u001b[0;32m    763\u001b[0m             \u001b[0mrefit_start_time\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtime\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mtime\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    764\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0my\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\.conda\\envs\\atml_202021\\lib\\site-packages\\sklearn\\utils\\validation.py\u001b[0m in \u001b[0;36minner_f\u001b[1;34m(*args, **kwargs)\u001b[0m\n\u001b[0;32m     70\u001b[0m                           FutureWarning)\n\u001b[0;32m     71\u001b[0m         \u001b[0mkwargs\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mupdate\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m{\u001b[0m\u001b[0mk\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0marg\u001b[0m \u001b[1;32mfor\u001b[0m \u001b[0mk\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0marg\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mzip\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msig\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mparameters\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0margs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m}\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 72\u001b[1;33m         \u001b[1;32mreturn\u001b[0m \u001b[0mf\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     73\u001b[0m     \u001b[1;32mreturn\u001b[0m \u001b[0minner_f\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     74\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\.conda\\envs\\atml_202021\\lib\\site-packages\\sklearn\\base.py\u001b[0m in \u001b[0;36mclone\u001b[1;34m(estimator, safe)\u001b[0m\n\u001b[0;32m     96\u001b[0m             raise RuntimeError('Cannot clone object %s, as the constructor '\n\u001b[0;32m     97\u001b[0m                                \u001b[1;34m'either does not set or modifies parameter %s'\u001b[0m \u001b[1;33m%\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 98\u001b[1;33m                                (estimator, name))\n\u001b[0m\u001b[0;32m     99\u001b[0m     \u001b[1;32mreturn\u001b[0m \u001b[0mnew_object\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    100\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mRuntimeError\u001b[0m: Cannot clone object <tensorflow.python.keras.wrappers.scikit_learn.KerasRegressor object at 0x000001C61600AD48>, as the constructor either does not set or modifies parameter learning_rate"
     ]
    }
   ],
   "source": [
    "from scipy.stats import reciprocal\n",
    "from sklearn.model_selection import RandomizedSearchCV\n",
    "\n",
    "param_distribs = {\n",
    "    \"n_hidden\": [(0), (1), (2), (3)],\n",
    "    \"n_neurons\": np.arange(1, 100),\n",
    "    \"learning_rate\": reciprocal(3e-4, 3e-2),\n",
    "}\n",
    "\n",
    "rnd_search_cv = RandomizedSearchCV(keras_reg, param_distribs, n_iter=10, cv=3, verbose=2)\n",
    "rnd_search_cv.fit(X_train, y_train, epochs=100,\n",
    "                  validation_data=(X_valid, y_valid),\n",
    "                  callbacks=[keras.callbacks.EarlyStopping(patience=10)])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The error above may be fixed in the next updates of the library."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'learning_rate': 0.0033625641252688094, 'n_hidden': 2, 'n_neurons': 42}"
      ]
     },
     "execution_count": 92,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rnd_search_cv.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "-0.3595289261637835"
      ]
     },
     "execution_count": 93,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rnd_search_cv.best_score_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = rnd_search_cv.best_estimator_.model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "rnd_search_cv.score(X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = rnd_search_cv.best_estimator_.model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.evaluate(X_test, y_test)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}