This repository houses coursework developed as a pair for the F21BSC Heriot-Watt course in biologically inspired computation.
Students:
The first part is an artifical neural network (ANN) written from scratch in Python and trained + tested against a prescribed dataset.
The ANN is a simple feedforward multilayer architecture with configurable hyperparameters (number of layers, neurons in each layer, etc.). It learns by means of backpropagation.
The dataset in question is the banknote authentication data described at: https://archive.ics.uci.edu/ml/datasets/banknote+authentication
The second part is a PSO implementation used to train the ANN architecture from part 1.