James Thornton

About

I joined Apple as a research scientist within the Machine Learning Research (MLR) group.

Prior to Apple, I did my PhD (DPhil) in the Computational Statistics and Machine Learning (CSML) research group within the Department of Statistics at the University of Oxford, supervised by George Deligiannidis and Arnaud Doucet.

My PhD thesis is available here.

Research Interests

I am interested in developing methodology at the intersection of: sampling methods, optimal transport and deep learning. I am particularly focused on dynamically evolving probability measures through filtering and diffusion models.

Notable work includes connecting diffusion models to optimal transport through the Diffusion Schrodinger Bridge, which was one of the first image-to-image diffusion models, its extension to the Riemannian setting; and Differentiable Particle Filtering, which provides a theoretically grounded approach to the long standing problem of parameter learning in state-space models. For an up-to-date list of works please see Google scholar.

I was one of the early advocates of diffusion models and maintain a website of diffusion/score papers: https://scorebasedgenerativemodeling.github.io/.

Before academia

Prior to Oxford, I worked in a modelling and analytics role for investment management firm, BlackRock. I completed my undergraduate and Master’s studies in Statistics at Warwick University where I worked with Anthony Lee on Monte Carlo methods for Bayesian non-parametric models.