Thomas Heap
My research focuses on ways of making probabilistic inference better.
In particular, I am currently working on ways of making wake-sleep type variational inference, fast maximum likelihood fitting of exponential family distributions and moment estimation scale better in the number of latent variables in the probabilistic model. In this research I am supervised by Laurence Aitchison.
In addition to this primary line of enquiry I am interested in:
- Deep learning theory
- Bayesian Machine learning more generally
Before starting my PhD I worked in industry, most recently as a research engineer working on speech recognition for fraud detection. Prior to that I did my MSc at The University of Edinburgh where my thesis was supervised by David Sterratt and Melanie Stefan.