Neural Network Multilayer Perceptron Single-Sample Training with Backpropagation Algorithm

This project implements the backpropagation (BP) algorithm for single-sample training in multilayer perceptron neural networks using MATLAB. The BP algorithm is a fundamental neural network training method that iteratively adjusts weights and biases to enhance network accuracy and performance. Our implementation includes key components such as: forward propagation computation using matrix operations, error calculation through target-output comparison, and backward propagation for gradient-based weight updates. The code structure involves defining network architecture (hidden layers, neurons), initializing parameters with random values, and implementing the training loop that processes one sample at a time. Key MATLAB functions utilized include matrix operations for efficient computation and custom functions for activation (e.g., sigmoid) and derivative calculations. This approach enhances the neural network's learning capability and prediction accuracy, making it suitable for various applications requiring adaptive machine learning models.