The goal was to create a nerual network from scratch that could classify a digit from an image. It is based on a paper I wrote in in 10th grade. I do not intend to maintain this project further.
This project uses the build-tool Maven to manage the one dependency. I tested it on Java 17 although it should work on all versions > 1.7. By default, the program includes a default network that worked quite well for me.
The program uses the csv format for its datasets, the first value represents the label of the following 784 grayscale pixels. In order to train the current nerual network I used the MNIST-Dataset as well as a self created one which turned out to work better. This might be caused by the fact, that the handwritten digits look quite diffrent from the mouse-drawn ones that this program creates.