standakozak / neural_network_visualization Goto Github PK

This project is a Tkinter GUI for creating and training MLP (Multi-Layer Perceptron) Neural Networks. It is inspired by Neural Networks and Deep Learning by Michael Nielsen and Sebastian Lague's amazing video on Neural Networks.

It enables creating neural networks with custom layer sizes and training it on Digits MNIST, Fashion MNIST and Quickdraw datasets with basic backpropagation algorithm.

In the Show Progress Screen, it monitors the accuracy and cost throughout multiple training epochs.

You can view the results of the last testing and see which test cases were problematic by clicking the Browse Outputs button. You can also test the network's performance on custom drawings (draw with left mouse button and erase with right button).