This project is entirely personal amusement. It resulted from an assignment to code a neural net and use SGD to recover a simple image which at the time didn't work out too well due to a lack of time. I am intending to implement most of the things in Goodfellow et al.'s Deep Learning and then maybe continue with something else.
- Started BatchNorm implementation, realized that current implementation of layers wouldn't work well with it. Have to separate linear and non-linear part and put the gradient descent algorithms maybe in the base module class.
- Added separate activation layer. Batch norm next.