This repository contains an implementation of a neural network built entirely from scratch using only NumPy and Python. The network is designed to classify handwritten digits from the MNIST dataset and consists of three layers. It achieves an accuracy of 85% on the training set and 87% on the test/validation set.
- 🛠️ From Scratch Implementation: No use of high-level libraries like TensorFlow or PyTorch, providing a deeper understanding of neural networks.
- 🏗️ Three-Layer Architecture: A simple yet effective neural network structure.
- 📈 High Accuracy: Achieves 85% accuracy on the training set and 87% on the test/validation set.
- 📚 Comprehensive Documentation: Detailed explanations of each step in the code.