This COMPENG 4SL4 assignment experiments with neural network classifiers with two hidden layers for binary classification. The banknote authentication data set is used. There are four predictor variables (i.e., features) and the goal is to predict if a banknote is authentic (class 0) or a forgery (class 1).
This code implements a neural network with forward and back propogation from scratch with ReLU activation for an arbitrary amount of hidden layers and nodes. Training with weight decay is also implemented as part of the assignment bonus.