Rust-Neural-Network

This project is a simple neural network implementation from scratch built in Rust, designed to work with the MNIST dataset of handwritten digits. The neural network achieves an accuracy of up to ~96.8% on the test set. The project was created to build a deeper understanding of neural networks.

A pre-trained model file, model.bin, is included in the repository. For optimal performance, compile the project in release mode, as it may run slowly in debug mode.

Getting started

git clone https://github.com/NLion74/Neural-Network-Scratch
cd Neural-Network-Scratch && cargo build --release
# Linux
chmod +x ./target/release/neural-network-scratch && ./target/release/neural-network-scratch
# Windows
.\target\release\neural-network-scratch.exe