My first foray into neural networks.
Implemented a neural network with various configurations. The best performance is given by MSE cost function with upto 96% accuracy on the test data on SGD. Over the course of implementing this algorithm, I've realised that SGD is much more faster than mini batch or batch gradient descent
Credits -
- MNSIT Loader from http://neuralnetworksanddeeplearning.com/chap1.html
- https://deeplearning.ai (Video lectures for deep learning)