This is an assignment work in PyTorch explaining the performance of CNN in MNIST Dataset.

There are two CNN networks which are modeled differently explaining the various network parameters.

Stochastic gradient descent is used to make the model convere at minima

CNN2 is modified to include batch normalisation which improved the performance by 4 percent.

It made us to learn a lot about Neural Networks and the various functions involved in it.

The output accuracy is around 94 percent.