Abstract:
Bayesian reasoning and inference play a prominent
role in machine learning (Andrieu et al., 2003). It
requires evaluation of integrals of the form EX (f ) =∫
Ω f (x)π(dx) that in most cases cannot been done a-
nalytically and are often high dimensional, the func-
tion f has multiple modes and probability distribu-
tion π has a heavy tail.