Developing tractable density estimation models


Theo Pavlakou

This PhD project will focus on developing density estimation models that are computationally tractable (i.e. their runtime and memory complexities are low order polynomials in the input). It will do so by considering autoregressive models, undirected graphical models and others. The aim is to develop models that are both good at generating new data from the learned probability distribution and that can give probabilities (densities) for test data points.


Supervisors: Iain Murray & Peter Richtarik