DeepNet

A project on understanding and implementing mathematics behind deep nets to build a neural network from scratch without external libraries. I have worked on MNIST_784 dataset by OpenML available in scikit-learn . I have done all the implemntations of forward propagation , back propagation , computing and minimizing cost , finetuning parameters and debugging in order to make this model 90% accurate. This neural network has 3 hidden layers and an output multiclass softmax layer , used relu in hidden layers .