Used the MNIST dataset for training and testing of various machine learning algorithms.

Coded softmax classifier using vanilla gradient descent in python. Accuracy = 90.99

Coded a Neural Net with 3 hidden layers from scratch in python. Implemented the forward pass and back-propogation algorithm using numpy. Trained the model for 50 iterations.
Accuracy = 95.39

The README inside the keras/ folder gives the full details of the experiment. The highlights are : best Accuracy = 99.31