Add RBFNN implementation with Iris classification

neural_network/rbfnn.py

@@ -0,0 +1,86 @@
+# RBF Neural Network (RBFNN) on Iris Dataset
+# ------------------------------------------------------
+# This script implements a Radial Basis Function Neural Network (RBFNN)
+# for classification tasks, using the Iris dataset as an example.
+#
+# Features:
+# - Uses KMeans to determine RBF centers.
+# - Applies Gaussian radial basis function as hidden layer activation.
+# - Trains output weights using least-squares fitting.
+#
+# Includes:
+# - Full training and prediction pipeline.
+# - Evaluation using classification accuracy.
+# ------------------------------------------------------
+
+import numpy as np
+from scipy.spatial.distance import cdist
+from sklearn.cluster import KMeans
+from sklearn.datasets import load_iris
+from sklearn.metrics import accuracy_score
+from sklearn.model_selection import train_test_split
+from sklearn.preprocessing import OneHotEncoder, StandardScaler
+
+
+class RBFNN:
+    def __init__(self, num_centers, gamma):
+        # Initialize with number of RBF centers and spread parameter (gamma)
+        self.num_centers = num_centers
+        self.gamma = gamma
+        self.centers = None
+        self.weights = None
+
+    def _rbf(self, x, centers):
+        # Compute Gaussian RBF activations for inputs x given the centers
+        dist = cdist(x, centers, "euclidean")  # Compute Euclidean distance to centers
+        return np.exp(-self.gamma * (dist**2))  # Apply Gaussian function
+
+    def train(self, x_data, y_data):
+        # Train the RBFNN
+        # Step 1: Use KMeans to find cluster centers for RBFs
+        kmeans = KMeans(n_clusters=self.num_centers, random_state=0).fit(x_data)
+        self.centers = kmeans.cluster_centers_
+
+        # Step 2: Compute RBF activations
+        rbf_activations = self._rbf(x_data, self.centers)
+
+        # Step 3: Solve output weights using least squares
+        self.weights = np.linalg.pinv(rbf_activations).dot(y_data)
+
+    def predict(self, x):
+        # Predict using learned weights and RBF activations
+        rbf_activations = self._rbf(x, self.centers)
+        return rbf_activations.dot(self.weights)
+
+
+if __name__ == "__main__":
+    # Load and preprocess Iris dataset
+    iris = load_iris()
+    x = iris.data  # Feature matrix
+    y = iris.target.reshape(-1, 1)  # Labels
+
+    # Standardize features
+    scaler = StandardScaler()
+    x_scaled = scaler.fit_transform(x)
+
+    # One-hot encode target labels for multi-class classification
+    encoder = OneHotEncoder(sparse_output=False)
+    y_encoded = encoder.fit_transform(y)
+
+    # Split data into training and testing sets
+    x_train, x_test, y_train, y_test = train_test_split(
+        x_scaled, y_encoded, test_size=0.2, random_state=42
+    )
+
+    # Initialize and train the RBF Neural Network
+    rbfnn = RBFNN(num_centers=10, gamma=1.0)
+    rbfnn.train(x_train, y_train)
+
+    # Predict on test set
+    y_pred_probs = rbfnn.predict(x_test)
+    y_pred = np.argmax(y_pred_probs, axis=1)
+    y_true = np.argmax(y_test, axis=1)
+
+    # Evaluate accuracy
+    accuracy = accuracy_score(y_true, y_pred)
+    print(f"Classification Accuracy: {accuracy:.4f}")