MOM_13_08_2019_Anil_Julien:
* Try tanh at the fc1 for better gradients and see the results(14th)
* Plot accuracies(14th)
* Plot the model results(later )
* Put the comments to all fun input and outputs(later)
* Do the hyper parameterization using hype(later)
* Try different distribution for trained linear layer(@home)