Re-enable five disabled algorithms and the perceptron (#15208)

Re-enable the four scikit-learn machine-learning examples and the
neural-network perceptron that had been disabled (renamed to
.broken.txt / .DISABLED), and modernize them so they import and run
cleanly on current scikit-learn and pass the doctest CI:

machine_learning/gaussian_naive_bayes.py
machine_learning/random_forest_classifier.py
  - Replace the removed sklearn.metrics.plot_confusion_matrix with
    ConfusionMatrixDisplay.from_estimator (removed in scikit-learn 1.2).
  - Drop the artificial time.sleep() calls.

machine_learning/gradient_boosting_regressor.py
machine_learning/random_forest_regressor.py
  - Replace the removed load_boston dataset (removed in scikit-learn
    1.2 for ethical reasons) with the bundled load_diabetes dataset so
    the examples run offline.
  - Avoid an unused-variable lint (RUF059).

neural_network/perceptron.py
  - Use a dedicated seeded random.Random instance instead of the global
    random state, so training is reproducible and thread-safe under the
    parallel test runner.
  - Cap training at epoch_number epochs so it always terminates even on
    non-linearly-separable data (previously an unbounded while True).
  - Have training() and sort() return their results instead of printing,
    per the contribution guidelines, and update the doctests accordingly.

Requested by @cclauss in #8029; perceptron follow-up to #15206.

Author: priya-sundaram-dev

…rning/gaussian_naive_bayes.py.broken.txt machine_learning/gaussian_naive_bayes.pymachine_learning/gaussian_naive_bayes.py.broken.txt renamed to machine_learning/gaussian_naive_bayes.py machine_learning/gaussian_naive_bayes.py.broken.txt renamed to machine_learning/gaussian_naive_bayes.py

@@ -1,20 +1,17 @@
 # Gaussian Naive Bayes Example
-import time
 
 from matplotlib import pyplot as plt
 from sklearn.datasets import load_iris
-from sklearn.metrics import accuracy_score, plot_confusion_matrix
+from sklearn.metrics import ConfusionMatrixDisplay, accuracy_score
 from sklearn.model_selection import train_test_split
 from sklearn.naive_bayes import GaussianNB
 
 
 def main():
-
     """
     Gaussian Naive Bayes Example using sklearn function.
     Iris type dataset is used to demonstrate algorithm.
     """
-
     # Load Iris dataset
     iris = load_iris()
 
@@ -27,23 +24,21 @@ def main():
 
     # Gaussian Naive Bayes
     nb_model = GaussianNB()
-    time.sleep(2.9)
-    model_fit = nb_model.fit(x_train, y_train)
-    y_pred = model_fit.predict(x_test)  # Predictions on the test set
+    nb_model.fit(x_train, y_train)
+    y_pred = nb_model.predict(x_test)  # Predictions on the test set
 
     # Display Confusion Matrix
-    plot_confusion_matrix(
+    ConfusionMatrixDisplay.from_estimator(
         nb_model,
         x_test,
         y_test,
         display_labels=iris["target_names"],
-        cmap="Blues",  # although, Greys_r has a better contrast...
+        cmap="Blues",
         normalize="true",
     )
     plt.title("Normalized Confusion Matrix - IRIS Dataset")
     plt.show()
 
-    time.sleep(1.8)
     final_accuracy = 100 * accuracy_score(y_true=y_test, y_pred=y_pred)
     print(f"The overall accuracy of the model is: {round(final_accuracy, 2)}%")
 

…radient_boosting_regressor.py.broken.txt …_learning/gradient_boosting_regressor.pymachine_learning/gradient_boosting_regressor.py.broken.txt renamed to machine_learning/gradient_boosting_regressor.py machine_learning/gradient_boosting_regressor.py.broken.txt renamed to machine_learning/gradient_boosting_regressor.py

@@ -1,33 +1,37 @@
 """Implementation of GradientBoostingRegressor in sklearn using the
-   boston dataset which is very popular for regression problem to
-   predict house price.
+diabetes dataset, a popular regression problem used to predict
+disease progression one year after baseline.
+
+Note: this example previously used the Boston house-price dataset,
+which was removed from scikit-learn (>=1.2) for ethical reasons.
+``load_diabetes`` is a drop-in bundled alternative that ships with
+scikit-learn, so the example runs offline.
 """
 
 import matplotlib.pyplot as plt
 import pandas as pd
-from sklearn.datasets import load_boston
+from sklearn.datasets import load_diabetes
 from sklearn.ensemble import GradientBoostingRegressor
 from sklearn.metrics import mean_squared_error, r2_score
 from sklearn.model_selection import train_test_split
 
 
 def main():
-
-    # loading the dataset from the sklearn
-    df = load_boston()
+    # loading the dataset from sklearn
+    df = load_diabetes()
     print(df.keys())
-    # now let construct a data frame
-    df_boston = pd.DataFrame(df.data, columns=df.feature_names)
-    # let add the target to the dataframe
-    df_boston["Price"] = df.target
+    # now let's construct a data frame
+    df_data = pd.DataFrame(df.data, columns=df.feature_names)
+    # let's add the target to the dataframe
+    df_data["Target"] = df.target
     # print the first five rows using the head function
-    print(df_boston.head())
+    print(df_data.head())
     # Summary statistics
-    print(df_boston.describe().T)
+    print(df_data.describe().T)
     # Feature selection
 
-    x = df_boston.iloc[:, :-1]
-    y = df_boston.iloc[:, -1]  # target variable
+    x = df_data.iloc[:, :-1]
+    y = df_data.iloc[:, -1]  # target variable
     # split the data with 75% train and 25% test sets.
     x_train, x_test, y_train, y_test = train_test_split(
         x, y, random_state=0, test_size=0.25
@@ -43,7 +47,7 @@ def main():
     test_score = model.score(x_test, y_test).round(3)
     print("Training score of GradientBoosting is :", training_score)
     print("The test score of GradientBoosting is :", test_score)
-    # Let us evaluation the model by finding the errors
+    # Let us evaluate the model by finding the errors
     y_pred = model.predict(x_test)
 
     # The mean squared error
@@ -52,7 +56,7 @@ def main():
     print(f"Test Variance score: {r2_score(y_test, y_pred):.2f}")
 
     # So let's run the model against the test data
-    fig, ax = plt.subplots()
+    _fig, ax = plt.subplots()
     ax.scatter(y_test, y_pred, edgecolors=(0, 0, 0))
     ax.plot([y_test.min(), y_test.max()], [y_test.min(), y_test.max()], "k--", lw=4)
     ax.set_xlabel("Actual")

…g/random_forest_classifier.py.broken.txt …ine_learning/random_forest_classifier.pymachine_learning/random_forest_classifier.py.broken.txt renamed to machine_learning/random_forest_classifier.py machine_learning/random_forest_classifier.py.broken.txt renamed to machine_learning/random_forest_classifier.py

@@ -1,18 +1,17 @@
 # Random Forest Classifier Example
+
 from matplotlib import pyplot as plt
 from sklearn.datasets import load_iris
 from sklearn.ensemble import RandomForestClassifier
-from sklearn.metrics import plot_confusion_matrix
+from sklearn.metrics import ConfusionMatrixDisplay
 from sklearn.model_selection import train_test_split
 
 
 def main():
-
     """
     Random Forest Classifier Example using sklearn function.
     Iris type dataset is used to demonstrate algorithm.
     """
-
     # Load Iris dataset
     iris = load_iris()
 
@@ -28,7 +27,7 @@ def main():
     rand_for.fit(x_train, y_train)
 
     # Display Confusion Matrix of Classifier
-    plot_confusion_matrix(
+    ConfusionMatrixDisplay.from_estimator(
         rand_for,
         x_test,
         y_test,

…ng/random_forest_regressor.py.broken.txt …hine_learning/random_forest_regressor.pymachine_learning/random_forest_regressor.py.broken.txt renamed to machine_learning/random_forest_regressor.py machine_learning/random_forest_regressor.py.broken.txt renamed to machine_learning/random_forest_regressor.py

@@ -1,24 +1,28 @@
 # Random Forest Regressor Example
-from sklearn.datasets import load_boston
+
+from sklearn.datasets import load_diabetes
 from sklearn.ensemble import RandomForestRegressor
 from sklearn.metrics import mean_absolute_error, mean_squared_error
 from sklearn.model_selection import train_test_split
 
 
 def main():
-
     """
     Random Forest Regressor Example using sklearn function.
-    Boston house price dataset is used to demonstrate the algorithm.
-    """
+    The diabetes dataset is used to demonstrate the algorithm.
 
-    # Load Boston house price dataset
-    boston = load_boston()
-    print(boston.keys())
+    Note: this example previously used the Boston house-price dataset,
+    which was removed from scikit-learn (>=1.2) for ethical reasons.
+    ``load_diabetes`` is a drop-in bundled alternative that ships with
+    scikit-learn, so the example runs offline.
+    """
+    # Load the diabetes dataset
+    diabetes = load_diabetes()
+    print(diabetes.keys())
 
     # Split dataset into train and test data
-    x = boston["data"]  # features
-    y = boston["target"]
+    x = diabetes["data"]  # features
+    y = diabetes["target"]
     x_train, x_test, y_train, y_test = train_test_split(
         x, y, test_size=0.3, random_state=1
     )

neural_network/perceptron.py.DISABLED neural_network/perceptron.pyneural_network/perceptron.py.DISABLED renamed to neural_network/perceptron.py

@@ -1,12 +1,15 @@
 """
-    Perceptron
-    w = w + N * (d(k) - y) * x(k)
+Perceptron
+w = w + N * (d(k) - y) * x(k)
 
-    Using perceptron network for oil analysis, with Measuring of 3 parameters
-    that represent chemical characteristics we can classify the oil, in p1 or p2
-    p1 = -1
-    p2 = 1
+Using perceptron network for oil analysis, with Measuring of 3 parameters
+that represent chemical characteristics we can classify the oil, in p1 or p2
+p1 = -1
+p2 = 1
+
+Reference: https://en.wikipedia.org/wiki/Perceptron
 """
+
 import random
 
 
@@ -18,6 +21,7 @@ def __init__(
         learning_rate: float = 0.01,
         epoch_number: int = 1000,
         bias: float = -1,
+        seed: int | None = 0,
     ) -> None:
         """
         Initializes a Perceptron network for oil analysis
@@ -26,6 +30,8 @@ def __init__(
         :param learning_rate: learning rate used in optimizing.
         :param epoch_number: number of epochs to train network on.
         :param bias: bias value for the network.
+        :param seed: seed for the (internal) random number generator so that
+            training is reproducible; pass ``None`` for non-deterministic weights.
 
         >>> p = Perceptron([], (0, 1, 2))
         Traceback (most recent call last):
@@ -54,29 +60,36 @@ def __init__(
         self.number_sample = len(sample)
         self.col_sample = len(sample[0])  # number of columns in dataset
         self.weight: list = []
+        # A dedicated RNG instance keeps training reproducible without touching
+        # the global ``random`` state (which other code/tests may rely on).
+        self._rng = random.Random(seed)
 
-    def training(self) -> None:
+    def training(self) -> int:
         """
-        Trains perceptron for epochs <= given number of epochs
-        :return: None
+        Trains the perceptron until it stops misclassifying the training data
+        or the maximum number of epochs (``epoch_number``) is reached, whichever
+        comes first. The epoch cap guarantees termination even if the data is
+        not linearly separable.
+
+        :return: the number of epochs the network was trained for.
+
         >>> data = [[2.0149, 0.6192, 10.9263]]
         >>> targets = [-1]
-        >>> perceptron = Perceptron(data,targets)
-        >>> perceptron.training() # doctest: +ELLIPSIS
-        ('\\nEpoch:\\n', ...)
-        ...
+        >>> perceptron = Perceptron(data, targets)
+        >>> perceptron.training()
+        5
         """
         for sample in self.sample:
             sample.insert(0, self.bias)
 
         for _ in range(self.col_sample):
-            self.weight.append(random.random())
+            self.weight.append(self._rng.random())
 
         self.weight.insert(0, self.bias)
 
         epoch_count = 0
 
-        while True:
+        while epoch_count < self.epoch_number:
             has_misclassified = False
             for i in range(self.number_sample):
                 u = 0
@@ -92,28 +105,28 @@ def training(self) -> None:
                             * self.sample[i][j]
                         )
                     has_misclassified = True
-            # print('Epoch: \n',epoch_count)
             epoch_count = epoch_count + 1
-            # if you want control the epoch or just by error
+            # stop early once every sample is classified correctly
             if not has_misclassified:
-                print(("\nEpoch:\n", epoch_count))
-                print("------------------------\n")
-                # if epoch_count > self.epoch_number or not error:
                 break
 
-    def sort(self, sample: list[float]) -> None:
+        return epoch_count
+
+    def sort(self, sample: list[float]) -> int:
         """
+        Classifies a single observation as P1 (-1) or P2 (1). The network must
+        be trained first.
+
         :param sample: example row to classify as P1 or P2
-        :return: None
+        :return: -1 if the sample is classified as P1, otherwise 1
+
         >>> data = [[2.0149, 0.6192, 10.9263]]
         >>> targets = [-1]
-        >>> perceptron = Perceptron(data,targets)
-        >>> perceptron.training() # doctest: +ELLIPSIS
-        ('\\nEpoch:\\n', ...)
-        ...
-        >>> perceptron.sort([-0.6508, 0.1097, 4.0009]) # doctest: +ELLIPSIS
-        ('Sample: ', ...)
-        classification: P...
+        >>> perceptron = Perceptron(data, targets)
+        >>> perceptron.training()
+        5
+        >>> perceptron.sort([2.0149, 0.6192, 10.9263])
+        -1
         """
         if len(self.sample) == 0:
             raise ValueError("Sample data can not be empty")
@@ -122,23 +135,16 @@ def sort(self, sample: list[float]) -> None:
         for i in range(self.col_sample + 1):
             u = u + self.weight[i] * sample[i]
 
-        y = self.sign(u)
-
-        if y == -1:
-            print(("Sample: ", sample))
-            print("classification: P1")
-        else:
-            print(("Sample: ", sample))
-            print("classification: P2")
+        return self.sign(u)
 
     def sign(self, u: float) -> int:
         """
         threshold function for classification
         :param u: input number
-        :return: 1 if the input is greater than 0, otherwise -1
-        >>> data = [[0],[-0.5],[0.5]]
-        >>> targets = [1,-1,1]
-        >>> perceptron = Perceptron(data,targets)
+        :return: 1 if the input is greater than or equal to 0, otherwise -1
+        >>> data = [[0], [-0.5], [0.5]]
+        >>> targets = [1, -1, 1]
+        >>> perceptron = Perceptron(data, targets)
         >>> perceptron.sign(0)
         1
         >>> perceptron.sign(-0.5)
@@ -224,8 +230,8 @@ def sign(self, u: float) -> int:
     network = Perceptron(
         sample=samples, target=target, learning_rate=0.01, epoch_number=1000, bias=-1
     )
-    network.training()
-    print("Finished training perceptron")
+    epochs = network.training()
+    print(f"Finished training perceptron in {epochs} epoch(s)")
     print("Enter values to predict or q to exit")
     while True:
         sample: list = []
@@ -235,4 +241,6 @@ def sign(self, u: float) -> int:
                 break
             observation = float(user_input)
             sample.insert(i, observation)
-        network.sort(sample)
+        classification = network.sort(sample)
+        label = "P1" if classification == -1 else "P2"
+        print(f"Sample: {sample} classification: {label}")