Implement stochastic gradient descent in your back propagation program
that you wrote in assignment 3. We will do the mini-batch SGD search. 

I. Mini-batch SGD algorithm:

Initialize random weights
for(k = 0 to n_epochs):
	Shuffle the rows (or row indices)
	for j = 0 to rows-1:
		Select the first k datapoints where k is the mini-batch size
		Determine gradient using just the selected k datapoints
		Update weights with gradient
	Recalculate objective

Your input, output, and command line parameters are the same as assignment 3.
We take the batch size k as input. We leave the offset for the final layer 
to be zero at this time.

Test your program on the XOR dataset:

1 0 0
1 1 1
-1 0 1
-1 1 0

1. Test your program on breast cancer and ionosphere given on the website. Is the 
mini-batch faster or the original one? How about accuracy?

2. Is the search faster or more accurate if you keep track of the best objective
in the inner loop?

Submit your assignments by copying your program to your AFS 
course folder /afs/cad/courses/ccs/s20/cs/677/002/<UCID>. The 
assignment is due on March 9th, 2020.