PDL

This project aims to connect further the parametric world of neural networks to the non-parametric world of GPs by advancing NNGPs.

This project aims to develop approximate MCMC methods for Bayesian deep learning in order to quantify the uncertainty of predictions made by neural networks.