This seminar is hosted by the Programme on Computational Cosmology, an Oxford Martin School Programme
Summary: Most data that we gather result from complex, hierarchical processes. However, parts of these processes may not be well-specified given the data, and our predictions may be over-confident and wrong if this uncertainty is not taken into account. In some scientific applications, expressing our uncertainty over different explanations of data is the whole goal of the analysis. The long-standing Markov chain Monte Carlo (MCMC) approach to inference explores different explanations of the data by a guided random walk.
Dr Iain Murray will illustrate the importance of probabilistic hierarchical models through two statistical problems: 1) Predicting and understanding basketball scores. 2) If you observe a snapshot of a dynamical system, such as the positions and instantaneous velocities of stars but no motion, can you infer the forces acting on the objects? He will then outline recent Markov chain Monte Carlo approaches for performing inference in these models.
Speaker: Dr Iain Murray, Lecturer in Machine Learning, School of Informatics, University of Edinburgh
Biography: Dr Murray is a lecturer in Machine Learning in the School of Informatics at the University of Edinburgh. His work spans a range of topics in statistics, in particular focussing on computational Bayesian inference as applied to everything from neuroscience to astronomy.
For a video introduction, see http://homepages.inf.ed.ac.uk/srodnes/CSE/?p=1935