Nick Whiteley
http://www.maths.bris.ac.uk/~manpw/
Wednesday 27th March 2013
Time: 4pm
Basement Seminar Room
Alexandra House, 17 Queen Square, London, WC1N 3AR
Particle Filters and Hidden Markov models
Since their appearance twenty years ago, particle filters have found thousands of applications across statistics, engineering, computer science and elsewhere, and they continue to feature at the core of cutting edge numerical methods. Over time, the complexity of models and the length of data records analyzed using particle filters has dramatically increased, from a few tens of data points to thousands. Moreover, there has been a shift in the functional ways that particle filters are deployed; historically the focus was on non-linear filtering, but the need to treat issues of model uncertainty in modern application areas sees particle filters increasingly used to calibrate and compare statistical models. This talk will give an overview of particle filters in this context, survey some theoretical properties which explain their practical utility and discuss some recent developments.