The purpose of this repository is to practice the understandings of different algorithms and to help others understand what is the blackbox made of.
Tested on some random points as a proof of concept.
Vanilla Neural Network with Tanh activation function and Softmax with cross entropy for final layer. Tested on MNIST dataset with 3 hidden layers and the layers dimensions are : Layer 1 (784,128) - Layer 2 (128,128) - Layer 3 (128,10). Achieved Testing accuracy : 80.61% with 2000 epochs