About
I am a Lecturer (equivalent to Assistant Professor) in the Department of Statistical Science at University College London, and a Group Leader at The Alan Turing Institute, the UK’s national institute for Data Science and AI, where I am affiliated to the Data-Centric Engineering programme. There, I lead research on the Fundamentals of Statistical Machine Learning with an emphasis on methodology for complex engineering models.
My research interests are at the interface of computational statistics, machine learning and applied mathematics. I work on methodology for statistical computation and inference for large scale and computationally expensive probabilistic models. In particular, I am interested in the development of algorithms for numerical integration and sampling, as well as methodology for inference with intractable models. In my research, I like to use a variety of tools including kernel methods, Monte Carlo methods, stochastic process theory and Stein’s method.
Prior to UCL, I was a PhD student on the joint centre for doctoral training between the Departments of Statistics at Warwick and Oxford, then spent a year first as research assistant in the Department of Mathematics at Imperial College London, then as a research associate in the Department of Engineering at the University of Cambridge.
For more details, see my Publications page or my Google Scholar profile. Alternatively, you can catch me at one of my Presentations. For supervision opportunities, see also this page.
News
I have two new papers on doing inference for intractable model. The first paper, called Minimum Stein Discrepancy Estimators, is a very flexible framework to do inference for models with unnormalised likelihoods which is grounded in Stein’s method and can recover many existing methods, such as score-matching and contrastive divergence, as special cases. See the following talk for more details. The second, called Statistical Inference for Generative Models with Maximum Mean Discrepancy considers the problem of inference for generative models, and studies the flexibility afforded by the choice of reproducing kernel.
Our paper on “Probabilistic Integration: A Role in Statistical Computation?” has been accepted for publication in Statistical Science and will appear together with invited discussions from leading researchers in statistics, as well as a rejoinder. There are three discussion pieces by (i) Fred Hickernel and R. Jagadeeswaran, (ii) Art Owen, and (iii) Michael Stein and Ying Hung.