In Graphical Models theory (part of Artificial Intelligence, networks are often constructed for which exact inference is intractable (NP-hard) and Monte-Carlo methods may be used instead.

Particle filtering is one such method, which approximates distributions by finite sets of weighted particles. The particle representation is then used with the Pearl message passing or junction tree algorithm. Particle filtering is especially useful in Dynamic Bayesian networks. It is also known as the Sequential Monte Carlo Method.

For examples of its use, see Ali Taylan Cemgil's PhD thesis, which applied particle filtering to musical state estimation.