mnist with different hyper parameters

In this repo, I trained a model on mnist data with different optimizers, loss functions and activation functions and make comparisons between results.

val accuracy

val accuracy zoomed

val loss

val loss zoomed

Tabel of accuracy and losses