The collapse of particle filters Peter J. Bickel Statistics Department University of California, Berkeley Abstract
Particle filters are Monte Carlo approximations to
the conditional distribution of the hidden process in HMM and state
space models. They have become very popular in situations where
the state space is large. Unfortunately it's been observed in
climate forecasting and other situations, where the dimension of the
state space is large that they can fail after 1 step of the
approximation. Using simple models I will try to explain why and
when this phenomenon can be expected.
This is joint work with Thomas Bengtsson and Bo Li. |