Torben Sell

Lecturer in Machine Learning

E-mail: torben.sell@ed.ac.uk

School of Mathematics
University of Edinburgh,
James Clerk Maxwell Building,
Peter Guthrie Tait Road,
Edinburgh, EH9 3FD, UK

Research interests: Statistical learning, Statistical methodology, Bayesian statistics, Bayesian neural networks, High-dimensional statistics, Markov chain Monte Carlo, Particle filtering, Reinforcement learning, Stochastic control, Function space sampling methods

I am currently also the principal investigator on an EPSRC Impact Acceleration Account (IAA) project on "How can machine learning and satellite data help with reducing costs in natural capital baselining and natural flood management?".

Please get in touch via email if you are interested in doing a PhD with me.


About

I am a lecturer in machine learning in the School of Mathematics, University of Edinburgh.

Before joining the School of Mathematics in Edinburgh, I was a PhD student at the Cantab Capital Institute for the Mathematics of Information (CCIMI) and the Statistical Laboratory within the Department of Pure Mathematics and Mathematical Statistics (DPMMS) at the University of Cambridge, supervised by Dr Sumeetpal S Singh. In April 2021, I moved to the University of Edinburgh with a LMS Early Career Fellowship, and became a postdoctoral research assistant working with Dr Tim I Cannings in May 2021. I started my current position in September 2023.


Publications

  1. Torben Sell, Thomas B. Berrett, Timothy I. Cannings, "Nonparametric classification with missing data", submitted, arXiv preprint.

  2. Torben Sell, Sumeetpal S Singh, "Trace-class Gaussian priors for Bayesian learning of neural networks with MCMC", Journal of the Royal Statistical Society Series B: Statistical Methodology, 2023, open access article.

  3. Chiara Toschi, Mona El-Sayed Hervig, Thiago Burghi, Torben Sell, Parisa Moazen, Li Huang, Ulrik Gether, Trevor W Robbins, Jeffrey W Dalley "Dissociating reward sensitivity and negative urgency effects on impulsivity in the 5-choice serial reaction time task", Brain and Neuroscience Advances, 2022, open access article.

  4. Jacob Vorstrup Goldman*, Torben Sell*, Sumeetpal S Singh, "Gradient-Based Markov Chain Monte Carlo for Bayesian Inference With Non-Differentiable Priors", Journal of the American Statistical Association, 2021, open access article. *equal contributions

  5. Torben Sell, "Advanced Bayesian Monte Carlo Methods for Inference and Control", PhD Thesis, 2021, pdf in the University of Cambridge repository.

  6. Torben Sell, "MCMC Methods for Functions: Modifying Old Algorithms to Make Them Faster", Essay for Part III of the Mathematical Tripos, pdf.

  7. Torben Sell, "A Convergence Analysis for Preconditioned Gradient Type Eigensolvers", Thesis for Bachelor of Science, pdf.

Teaching

I am course organiser for Statistical Research Skills, taught in Semester 2. In the past, I taught the following courses:

Academic Activities

I am a member of the Royal Statistical Society, the International Society for Bayesian Analysis and the Institute for Mathematical Statistics.

Since September 2021, I am one of the two organisers of the Statistics Seminar Series at the University of Edinburgh.

In 2020 I co-organised the Mathematics of Data Science conference, and wrote a short lessons learned paragraph about it.

In the academic year 2018/2019 I was one of two directors of the Part III seminar series at the University of Cambridge.

I enjoy reviewing papers within my area of expertise.

Miscellaneous

I love sailing and wrote a (hopefully) entertaining logbook entry about the first half of my summer sails 2022 and a logbook ballad about my summer sail 2023.