XClose

Statistical Science

Home
Menu

Dr Samuel Livingstone

PositionLecturer
Phone (external)+44(0)20 7679 1858
Phone (internal)41858
Email(*)samuel.livingstone
Personal webpagesamueljlivingstone.wixsite.com/webpage
ThemesComputational Statistics, General Theory and MethodologyStochastic Modelling and Time Series

* @ucl.ac.uk

Biographical Details

I have been a lecturer in Statistics at UCL since January 2018. Before that I was a postdoctoral researcher with Christophe Andrieu at the University of Bristol, as part of the 'i-like: intractable likelihood' project, and before that I was a PhD student at UCL, supervised by Mark Girolami and Alex Beskos. In a previous life I worked as a transport planner specialising in demand and revenue forecasting.

Research Interests

My primary research area is Bayesian computation, with a particular focus on Monte Carlo methods. I am also interested privacy-preserving statistical methods, and the general development of mathematical/statistical models for interesting applications.

Selected publications

  • The Geometric Foundations of Hamiltonian Monte Carlo, with Michael Betancourt, Simon Byrne and Mark Girolami, Bernoulli 23 (4A), 2257-2298.
  • Gradient-free Hamiltonian Monte Carlo with efficient kernel exponential families, with Heiko Strathmann, Dino Sejdinovic, Zoltan Szabo and Arthur Gretton, Advances in Neural Information Processing Systems, 955-963.
  • Information-geometric Markov chain Monte Carlo methods using diffusions, with Mark Girolami, Entropy 16 (6), 3074-3102.
  • Langevin diffusions and the Metropolis-adjusted Langevin algorithm, with T Xifara, C Sherlock, S Byrne and M Girolami Statistics & Probability Letters 91, 14-19
  • Kinetic energy choice in Hamiltonian/hybrid Monte Carlo, with Michael Faulkner and Gareth Roberts, submitted.
  • On the geometric ergodicity of Hamiltonian Monte Carlo, with Michael Betancourt, Simon Byrne and Mark Girolami, submitted.
  • Geometric ergodicity of the Random Walk Metropolis with position-dependent proposal covariance, submitted.