Computational Statistics and Machine Learning

The members of the Computational Statistics and Machine Learning Group (OxCSML) have research interests spanning Statistical Machine Learning, Monte Carlo Methods and Computational Statistics, Statistical Methodology and Applied Statistics.

Research in Statistical Machine Learning focuses on Bayesian inference for complex non-parametric models and complements research in Monte Carlo methods for the same class of problems. Researchers in Statistical Methods develop very general statistical methodology. Research in Applied Statistics motivates the more theoretical work in this group and some staff focus on developing statistical methodology ‘on demand’ in a wide range of application domains.

Research highlights include:

Professor Doucet and collaborators’ new Monte Carlo methods for inference using particle Markov chain Monte Carlo (a JRSSB read paper 2010)

Professor Teh and collaborators’ development of “sequence memoizer” statistical models for language, based on hierarchical Bayesian nonparametrics (best paper, 2009 International Conference of Artificial Intelligence and Statistics, invited paper Communications of the ACM 2011)

Professor Holmes and collaborators’ use of multi-core graphics processing units (GPUs) for advanced Monte Carlo methods (JCGS 2010, named by journal one of three "Highlights of the Year")

Professor Nicholls and collaborators’ model for the phylogeny of binary traits as a branching birth death process on sets, together with its validated application to the evolution of human vocabulary (JRSSB 2008, JRSSC 2011).

Academic Staff

 

Research staff

  • Angelos Armen
  • Dr Louis Aslett
  • Marco Battiston
  • Dr Seth Flaxman
  • Luke Kelly
  • Dr Tigran Nagapetyan
  • Dr Konstantina Palla
  • Dr Chieh-Hsi Wu

Links

  • Computational Statistics and Machine Learning (OxCSML) website
  • Oxford Sparks animation - What is Machine Learning?
  • Link to Professor Robin Evans' blog It's a Stats Life
  • Intractable likelihood - Link to i-like project