Running:

Please refer test.py

config.py:

Contains the Fixed Parameters for the network such as input Dimension, Number of Classes, Initial bias etc

activations.py:

Contains the Forward and Backward Pass Code for Activation Functions : ['relu', 'tanh']

test.py:

Testing and Training Script
Gradient Method Options: ['mbsgd', 'adagrad', 'gdm']
mbsgd: Mini Batch Stochastic Gradient Descent
adagrad: AdaGrad
gdm: Gradient Descent with Momentum

network.py:

init : Initialises the neurons, Weight values, Gradients [0], Directions for gdm, Past Gradient Squared Sum for Adagrad
forward: Forward Pass Code
predict: Predicts the Class given an input vector
update: Given a Method updates the weights of the network using gradients etc
backward: Backward pass (Accumulates Gradients )
trloss: return the loss for a training example
train : training function given training and validation data

Other Information:

Loss Function: Cross Entropy Loss
Default training method: Mini Batch Stochastic gradient Descent