Master's Project - working on PriorVAE

Open source code for my Master's project

The code underpinning my Master’s Project at Oxford with Seth Flaxman and Elizaveta Semenova, extending and building on some of their prior work, a GP approximation called PriorVAE for much, much faster MCMC. My project was to explore how we might train a VAE to better approximate the GP by changing the loss function - my primary explorations were with the maximum mean discrepancy. The project was also a good opportunity to reimplement PriorVAE and get more familiar with (the pains of) working with the computing cluster, which ran SLURM.