RBF Neural Network Classifier with 4-Input Layer, 3-Hidden Layer, and 2-Output Layer Architecture

This documentation presents an RBF neural network classifier implemented in MATLAB, designed with a specific architecture comprising 4 input neurons, 3 hidden layers with radial basis functions, and 2 output nodes for binary classification tasks. The implementation includes comprehensive training and testing datasets to facilitate model evaluation and performance validation. The classifier employs Gaussian radial basis functions in the hidden layers, utilizing the Euclidean distance metric between input vectors and cluster centers. Key MATLAB functions implemented include k-means clustering for center initialization, variance calculation for basis function widths, and linear regression for output weight optimization. The training process follows a two-phase approach: unsupervised clustering for hidden layer parameters followed by supervised learning for output layer weights. This program enables effective classification and prediction of input data patterns, serving as both a practical tool for pattern recognition tasks and an educational resource for understanding neural network principles. The code structure allows for straightforward modification of network parameters, including adjustment of hidden layer neurons and learning rates for customized applications.