Digit Recognition via Perceptrons: Convergence and Robustness Analysis.

This project implements and evaluates a Perceptron neural network model for recognizing digit patterns (0-9) represented in 7x4 matrices. It explores two critical stages: training, where convergence speed is analyzed under various learning rates, and validation, where the model's generalization capability is tested against noisy data (20%). The results demonstrate that the Perceptron achieves 100% training accuracy, though its performance significantly degrades when faced with random noise.