Created mlp_activation_comparison.py

machine_learning/mlp_activation_comparison.py

@@ -0,0 +1,57 @@
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.datasets import make_moons
+from sklearn.model_selection import train_test_split
+from sklearn.neural_network import MLPClassifier
+
+
+def compare_activations() -> None:
+    """
+    Demonstrates the effect of different activation functions on a simple dataset
+    using scikit-learn's MLPClassifier.
+
+    >>> compare_activations()  # doctest: +SKIP
+    This function trains models and plots decision boundaries for each activation.
+    """
+    x, y = make_moons(n_samples=200, noise=0.25, random_state=3)
+    x_train, x_test, y_train, y_test = train_test_split(
+        x, y, stratify=y, random_state=42
+    )
+
+    activations = ["identity", "logistic", "tanh", "relu"]
+
+    for activation in activations:
+        mlp = MLPClassifier(
+            hidden_layer_sizes=[50],
+            max_iter=1000,
+            activation=activation,
+            random_state=0,
+        )
+        mlp.fit(x_train, y_train)
+
+        print(
+            f"Activation: {activation}, "
+            f"Train Accuracy: {mlp.score(x_train, y_train):.2f}, "
+            f"Test Accuracy: {mlp.score(x_test, y_test):.2f}"
+        )
+
+        # Decision boundary
+        x_min, x_max = x[:, 0].min() - 0.5, x[:, 0].max() + 0.5
+        y_min, y_max = x[:, 1].min() - 0.5, x[:, 1].max() + 0.5
+        xx, yy = np.meshgrid(
+            np.linspace(x_min, x_max, 200),
+            np.linspace(y_min, y_max, 200),
+        )
+        z = mlp.predict(np.c_[xx.ravel(), yy.ravel()])
+        z = z.reshape(xx.shape)
+
+        plt.contourf(xx, yy, z, alpha=0.3)
+        plt.scatter(x_train[:, 0], x_train[:, 1], c=y_train, marker="o", label="Train")
+        plt.scatter(x_test[:, 0], x_test[:, 1], c=y_test, marker="s", label="Test")
+        plt.title(f"Activation: {activation}")
+        plt.legend()
+        plt.show()
+
+
+if __name__ == "__main__":
+    compare_activations()