William finishes his masters degree

William Pugsley finished his MMath degree in Computational Mathematics. He was jointly supervised by Jeff Orchard (CS) and Roxane Itier (Psychology). His research was on a neural network model of human face perception.

Abstract

Predictive coding is a theory of neural computation inspired by biological evidence and with mathematical backing. As with any theory, it is only useful if it is explanatory and predictive. This project uses predictive coding neural networks to model the facial perception pathway in human brains. We accurately reproduced the results of an electroencephalography experiment, which was the motivating study for this project [38]. In that study, the authors found a delay in P1, N170, and P2 Event Related Potentials (ERP) components in the EEG recordings of participants when shown faces with fewer parafoveal features. Our predictive coding model exhibits the same time delay. We also propose a mechanistic explanation for this behaviour based on the activities of neurons in V1, the inferior occipital gyrus, and the posterior fusiform gyrus.

You can read his report here.