Sequential Monte Carlo

Sequential Monte Carlo (SMC) is a family of sampling methods for approximating sequences of distributions, particularly useful in state-space models and time-series analysis. SMC maintains a population of weighted particles that are propagated, weighted, and resampled as new data arrives. Unlike MCMC, which requires convergence at each step, SMC advances the particle cloud sequentially, making it natural for online inference and filtering.

The method generalizes the particle filter to arbitrary proposal distributions and arbitrary target sequences. It is widely used in robotics, econometrics, and computational biology.

See also: Particle filter, Markov Chain Monte Carlo, Approximate inference, State-space model