Visualising Topological Structures of Activation in Artificial Neural Networks

This project report describes a first approach to creating a visualisation of an artificial neural network, that visualises the topology of the network given an individual data input that the network has learned to recognise. A survey of previous attempts to visualise both artificial and biological neural networks is presented, as well as a survey of various techniques used in other forms of network visualisation that could be applied to visualising artificial neural networks. This is followed by a detailed description of the method implemented in this project, followed by results from the visualisation