A simple yet powerful neural network model that recognizes handwritten digits (0–9) using a Multilayer Perceptron (MLP) built from scratch in Python.
This project demonstrates how neural networks learn to classify images without using high-level libraries like TensorFlow or PyTorch.
This project implements a feedforward neural network to classify digits from the MNIST dataset — a classic benchmark in computer vision.
The model is built from scratch, meaning the forward propagation, activation functions, backpropagation, and weight updates are implemented manually using only NumPy.
- Implementation of Multilayer Perceptron (MLP) using NumPy
- Forward & Backward propagation coded manually
- Supports mini-batch gradient descent
- Visualization of accuracy and loss curves
- Evaluation on MNIST test set