Basic multi-layer perceptron model in Rust trained on the MNIST dataset
To use the MNIST Model, just build and run the project with cargo run --release and head to http://localhost:80, there you can start training the model.
By increasing the batch size continusly from 5 to 256 within about 10-15 runs and at the same time slowly lowering the learning rate from about 2.5 to 1.5, over 96% Accuracy on the test data have be achieved.
Have fun!
