|
Abstract:
Particle filters allow us to carry out on-line
approximation of probability distributions using samples (particles). They
are very useful in scenarios involving real-time signal processing, where
data arrival is inherently sequential. Computational simplicity in the
form of not having to store all the data might also constitute an additional
motivating factor for these methods. My talk will introduce these methods
and discuss variance reduction techniques (Rao-Blackwellisation), on-line
model selection and parameter estimation, integration with MCMC methods,
and application to dynamic Bayesian networks, on-line semi-parametric regression
and probit classification.