Sekvenční Monte Carlo metody

Abstract

Monte Carlo methods provide recipe to convert numerical problems onto problems of random variable expected value estimation. Numerical solvers approximate distribution of given random variable to provide demanded estimate. Distribution approximations may be obtained by the means of Monte Carlo algorithms. Their subclass, Sequential Monte Carlo algorithms, are used for approximating sequences of distributions. Such needs arise from Bayesian data analysis. We will briefly introduce Bayesian inference and consequent need to approximate so-called aposteriori distributions. Further, we will state common properties of sequential methods, basic algorithms and their improvements and a couple of examples arising in Bayesian inference, on which we will demonstrate described algorithms.