Light-MNIST.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "# Imports\n",
    "from __future__ import print_function\n",
    "from matplotlib import pyplot as plt\n",
    "import keras\n",
    "from keras.datasets import mnist\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Flatten\n",
    "from keras.layers import Conv2D, MaxPooling2D\n",
    "from IPython.display import clear_output\n",
    "from keras import backend as K"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Load MNIST data\n",
    "(x_train, y_train), (x_test, y_test) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of training data (60000, 28, 28)\n",
      "Shape of labels (60000,)\n"
     ]
    }
   ],
   "source": [
    "print(\"Shape of training data\", x_train.shape)\n",
    "print(\"Shape of labels\", y_train.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nDwRF+IGg/g9u3HZr7xcc8gAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fef9c089150>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize Data\n",
    "plt.imshow(x_train[1])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Model Parameters\n",
    "batch_size = 128\n",
    "num_classes = 10\n",
    "epochs = 25\n",
    "\n",
    "# input image dimensions\n",
    "img_rows, img_cols = 28, 28"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Convert images into proper format that keras reads\n",
    "if K.image_data_format() == 'channels_first':\n",
    "    x_train = x_train.reshape(x_train.shape[0], 1, img_rows, img_cols)\n",
    "    x_test = x_test.reshape(x_test.shape[0], 1, img_rows, img_cols)\n",
    "    input_shape = (1, img_rows, img_cols)\n",
    "else:\n",
    "    x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1)\n",
    "    x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)\n",
    "    input_shape = (img_rows, img_cols, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Data Normalization\n",
    "x_train = x_train.astype('float32')\n",
    "x_test = x_test.astype('float32')\n",
    "x_train /= 255\n",
    "x_test /= 255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# convert class vectors to binary class matrices\n",
    "y_train = keras.utils.to_categorical(y_train, num_classes)\n",
    "y_test = keras.utils.to_categorical(y_test, num_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Plot Loss in real time -- Ref [stared]\n",
    "\n",
    "class PlotLosses(keras.callbacks.Callback):\n",
    "    def on_train_begin(self, logs={}):\n",
    "        self.i = 0\n",
    "        self.x = []\n",
    "        self.losses = []\n",
    "        self.val_losses = []\n",
    "        \n",
    "        self.fig = plt.figure()\n",
    "        \n",
    "        self.logs = []\n",
    "\n",
    "    def on_epoch_end(self, epoch, logs={}):\n",
    "        \n",
    "        self.logs.append(logs)\n",
    "        self.x.append(self.i)\n",
    "        self.losses.append(logs.get('loss'))\n",
    "        self.val_losses.append(logs.get('val_loss'))\n",
    "        self.i += 1\n",
    "        clear_output(wait=True)\n",
    "        plt.plot(self.x, self.losses, label=\"loss\")\n",
    "        plt.plot(self.x, self.val_losses, label=\"val_loss\")\n",
    "        plt.legend()\n",
    "        plt.show();\n",
    "        \n",
    "plot_losses = PlotLosses()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Baseline Model Definition\n",
    "model = Sequential()\n",
    "model.add(Conv2D(32, kernel_size=(5, 5),\n",
    "                 activation='relu',\n",
    "                 input_shape=input_shape,\n",
    "                 padding='same'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(Conv2D(64, (5, 5), activation='relu', padding='same'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(Flatten())\n",
    "model.add(Dense(1024, activation='relu'))\n",
    "model.add(Dense(num_classes, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 28, 28, 32)        832       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 14, 14, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 14, 14, 64)        51264     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 7, 7, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 3136)              0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1024)              3212288   \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                10250     \n",
      "=================================================================\n",
      "Total params: 3,274,634\n",
      "Trainable params: 3,274,634\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Model Summary\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX4AAAD8CAYAAABw1c+bAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nZaYAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fef5875a350>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "60000/60000 [==============================] - 5s - loss: 0.0043 - acc: 0.9988 - val_loss: 0.0341 - val_acc: 0.9927\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7fef99f1d890>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train Model\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer=keras.optimizers.Adam(),\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "model.fit(x_train, y_train,\n",
    "          batch_size=batch_size,\n",
    "          epochs=epochs,\n",
    "          verbose=1,\n",
    "          validation_data=(x_test, y_test), \n",
    "          callbacks=[plot_losses])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Varying the FC layer to reduce num of parameters\n",
    "model = Sequential()\n",
    "model.add(Conv2D(32, kernel_size=(5, 5),\n",
    "                 activation='relu',\n",
    "                 input_shape=input_shape,\n",
    "                 padding='same'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(Conv2D(64, (5, 5), activation='relu', padding='same'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(Flatten())\n",
    "model.add(Dense(128, activation='relu'))\n",
    "model.add(Dense(num_classes, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_3 (Conv2D)            (None, 28, 28, 32)        832       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 14, 14, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 14, 14, 64)        51264     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 7, 7, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten_2 (Flatten)          (None, 3136)              0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 128)               401536    \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 10)                1290      \n",
      "=================================================================\n",
      "Total params: 454,922\n",
      "Trainable params: 454,922\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Model Summary\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYQAAAD8CAYAAAB3u9PLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fef3413a0d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "60000/60000 [==============================] - 4s - loss: 2.8163e-04 - acc: 1.0000 - val_loss: 0.0332 - val_acc: 0.9938\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7fef9998bfd0>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train Model\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer=keras.optimizers.Adam(),\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "model.fit(x_train, y_train,\n",
    "          batch_size=batch_size,\n",
    "          epochs=epochs,\n",
    "          verbose=1,\n",
    "          validation_data=(x_test, y_test), \n",
    "          callbacks=[plot_losses])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "### Model still achieves 99.38 accuracy, with 14% the numbers of trainable parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# What can we do with just 1 Conv Layer\n",
    "model = Sequential()\n",
    "model.add(Conv2D(2, kernel_size=(5, 5),\n",
    "                 activation='relu',\n",
    "                 input_shape=input_shape,\n",
    "                 padding='same'))\n",
    "model.add(MaxPooling2D(pool_size=(4, 4)))\n",
    "model.add(Flatten())\n",
    "model.add(Dense(128, activation='relu'))\n",
    "model.add(Dense(num_classes, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_5 (Conv2D)            (None, 28, 28, 2)         52        \n",
      "_________________________________________________________________\n",
      "max_pooling2d_5 (MaxPooling2 (None, 7, 7, 2)           0         \n",
      "_________________________________________________________________\n",
      "flatten_3 (Flatten)          (None, 98)                0         \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 128)               12672     \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 10)                1290      \n",
      "=================================================================\n",
      "Total params: 14,014\n",
      "Trainable params: 14,014\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Model Summary\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAD8CAYAAAB9y7/cAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nFPAiIhlKAS8ikqH+PyDCuiDLPHdXAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fef7c1e8150>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "60000/60000 [==============================] - 1s - loss: 0.0377 - acc: 0.9876 - val_loss: 0.0724 - val_acc: 0.9786\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7fef58351b50>"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train Model\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer=keras.optimizers.Adam(),\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "model.fit(x_train, y_train,\n",
    "          batch_size=batch_size,\n",
    "          epochs=epochs,\n",
    "          verbose=1,\n",
    "          validation_data=(x_test, y_test), \n",
    "          callbacks=[plot_losses])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "### Model still achieves 97.86 accuracy, with 0.4% the numbers of trainable parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 1x1 conv layer\n",
    "model = Sequential()\n",
    "model.add(Conv2D(2, kernel_size=(3, 3),\n",
    "                 activation='relu',\n",
    "                 input_shape=input_shape,\n",
    "                 padding='valid'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(Conv2D(2, (1, 1), activation='relu', padding='valid'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(Flatten())\n",
    "model.add(Dense(64, activation='relu'))\n",
    "model.add(Dense(num_classes, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_6 (Conv2D)            (None, 26, 26, 2)         20        \n",
      "_________________________________________________________________\n",
      "max_pooling2d_6 (MaxPooling2 (None, 13, 13, 2)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 13, 13, 2)         6         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_7 (MaxPooling2 (None, 6, 6, 2)           0         \n",
      "_________________________________________________________________\n",
      "flatten_4 (Flatten)          (None, 72)                0         \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 64)                4672      \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 10)                650       \n",
      "=================================================================\n",
      "Total params: 5,348\n",
      "Trainable params: 5,348\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Model Summary\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAD8CAYAAAB9y7/cAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fef15dfe650>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "60000/60000 [==============================] - 2s - loss: 0.0740 - acc: 0.9760 - val_loss: 0.1004 - val_acc: 0.9675\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7fef58081490>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Train Model\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer=keras.optimizers.Adam(),\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "model.fit(x_train, y_train,\n",
    "          batch_size=batch_size,\n",
    "          epochs=epochs,\n",
    "          verbose=1,\n",
    "          validation_data=(x_test, y_test), \n",
    "          callbacks=[plot_losses])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "### Model still achieves 96.75% accuracy, with 0.15% the numbers of trainable parameters"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}