Master Class: Particle Filtering and Smoothing for State-Space Models by Arnaud Doucet

3rd October 2012, 1pm, UCL

To book please go to: http://csmlmasterclassdoucet.eventbrite.com/

State-space models are a popular class of time series models which are ubiquitous in econometrics, ecology, robotics, signal processing, statistics etc. Beyond finite state-space and linear Gaussian models, approximate inference in state-space models relies either on analytical or numerical approximations of the posterior distributions of interest. Particle methods are a class of sequential Monte Carlo methods which are flexible, easily parallelizable and provide consistent estimates. In this talk, I will review standard and advanced particle filtering and smoothing techniques. I will also discuss theoretical results which shed light on the performance of these approaches.

This is the first of three talks. There is no need to sign up for the talks on 4 and 5 October (same time and location). For more information please see: http://www.csml.ucl.ac.uk/events/series/10

Lunch will follow each lecture, in the Roberts Building Foyer (G02)

Arnaud will be available for individual of group meetings during the day from Monday 1 to Friday 5 October 2012. Please email Victoria Nicholl v.nicholl@ucl.ac.uk

Distinguished Lecture: Maximum Likelihood Particle Parameter Estimation for State-Space Models
Thursday 4 October 2012
Talk: Roberts G06 Sir Ambrose Fleming LT – 13:00-14:00
Lunch: Roberts Foyer G02 – 14:00-15:00

Distinguished Lecture: Bayesian Parameter Inference in State-Space Models using Particle Markov chain Monte Car
Friday 5 October 2012
Talk: Roberts G06 Sir Ambrose Fleming LT – 13:00-14:00
Lunch: Roberts Foyer G02 – 14:00-15:00

Sponsored by DeepMind Technologies: an ambitious London-based startup building general-pupose learning algorithms, with initial product applications in mobile social gaming.