Project for Statistical Machine Learning:

We seek to test different regularizers(L2,Dropout,Batchnorm) and their combinations against each other to study the effects each of them have. To do so, we shall be training a convolutional neural network implemented in the keras backend found in tensorflow 2, with the datasets taken from the same implementation. The data we shall be analysing is taken from the MNIST and CIFAR10 or CIFAR100 datasets.

The Proposal File Attached goes through the outline of the project. All code shall be saved in the Code folder, which includes a data subdirectory. This subdirectory will be empty on the git and is solely used to include the data in.