Dr Ben Calderhead
|Themes||Computational Statistics, General Theory and Methodology|
Ben graduated in Mathematics from the University of Glasgow, during which time he also studied Pure Mathematics for one year at the University of Mainz (Germany). He then completed a Masters (by research) in Computational Statistics, before completing a doctorate in Computational Statistics as a Microsoft Research European PhD Scholar, also at the University of Glasgow. Over the last 8 years, Ben has completed a number of business internships including software development at Data Connection (now Metaswitch Networks), business analysis at Procter & Gamble, and financial analysis in M&A at Lazard. He also ran his own computer company and shop for 3 years while at school.
He now works as a Research Fellow within CoMPLEX (Centre for Mathematics and Physics in the Life Sciences and EXperimental Biology) and the Department of Statistical Science at University College London.
My research interests lie in the development of general Bayesian statistical methodology, with applications for quantifying uncertainty and reasoning about hypothesised mechanisms in physical systems. My current work focuses on models defined using nonlinear differential equations and aggregated continuous-time Markov processes.
Throughout my current Research Fellowship I shall be developing and applying efficient Bayesian inferential methods that exploit the geometry of complex statistical models, in order to address the main challenges of parameter identifiability, multimodality, and in particular systematic model comparison in a hypothesis driven manner. This research is carried out as part of the interdisciplinary EPSRC project 2020 Science, which involves research groups from UCL, Oxford University and Microsoft Research.
My previous work includes development of novel Markov chain Monte Carlo methodology that exploits the natural Riemannian geometry of a statistical model. In my doctoral thesis I demonstrated how to derive generalisations of the Metropolis-adjusted Langevin algorithm and the Hybrid Monte Carlo algorithm defined on a Riemannian manifold. The resulting algorithms allow for efficient Bayesian statistical inference over many classes of statistical models and resolve many shortcomings of existing Monte Carlo algorithms when sampling from target densities that may be high dimensional and exhibit strong correlation structure. In particular I considered examples of Bayesian inference on logistic regression models, log-Gaussian Cox point process models, stochastic volatility models, and both parameter and model level inference of dynamical systems described by nonlinear differential equations.
- M. Girolami and B. Calderhead, Riemann Manifold Langevin and Hamiltonian Monte Carlo Methods (with Discussion), Journal of the Royal Statistical Society: Series B, Vol. 73 (2), 2011
- B. Calderhead & M. Girolami, Statistical analysis of nonlinear dynamical systems using differential geometric sampling methods, Journal of the Royal Society Interface Focus, Vol. 1 (6), 821-835, 2011
- L. Mohamed, B. Calderhead, M. Filippone, M. Christie & M. Girolami, Population MCMC methods for history matching and uncertainty quantification, Journal of Computational Geosciences, Vol. 16 (2), 423-436, 2012
- B. Calderhead & M. Girolami, Estimating Bayes Factors via Thermodynamic Integration and Population MCMC, Computational Statistics and Data Analysis, Vol. 53, 4028-4045, 2009
- B. Calderhead, M. Girolami & N. Lawrence, Accelerating Bayesian Inference over Nonlinear Differential Equations with Gaussian Processes, Advances in Neural Information Processing Systems, Vol. 21, 217-224, MIT Press, 2009
- M. Girolami, B. Calderhead & V. Vyshemirsky, System Identification and Model Ranking: The Bayesian Perspective, Learning and Inference in Computational Systems Biology, MIT Press, 2010
- X. Luo, B. Calderhead, H. Liu & W. Li, On the initial configurations of collapsible tube flow, Computers and Structures, Elsevier Press, Vol. 85, 977-987, 2007
Page last modified on 14 dec 12 11:23