An Adaptive Sequential Monte Carlo Sampler - Statistics > ComputationReportar como inadecuado

An Adaptive Sequential Monte Carlo Sampler - Statistics > Computation - Descarga este documento en PDF. Documentación en PDF para descargar gratis. Disponible también para leer online.

Abstract: Sequential Monte Carlo SMC methods are not only a popular tool in theanalysis of state space models, but offer an alternative to MCMC in situationswhere Bayesian inference must proceed via simulation. This paper introduces anew SMC method that uses adaptive MCMC kernels for particle dynamics. Theproposed algorithm features an online stochastic optimization procedure toselect the best MCMC kernel and simultaneously learn optimal tuning parameters.Theoretical results are presented that justify the approach and give guidanceon how it should be implemented. Empirical results, based on analysing datafrom mixture models, show that the new adaptive SMC algorithm ASMC can bothchoose the best MCMC kernel, and learn an appropriate scaling for it. ASMC witha choice between kernels outperformed the adaptive MCMC algorithm of Haario etal. 1998 in 5 out of the 6 cases considered.

Autor: Paul Fearnhead, Benjamin M. Taylor


Documentos relacionados