Output-Sensitive Adaptive Metropolis-Hastings for Probabilistic Programs
Artificial Intelligence
2015-05-07 v2 Machine Learning
Abstract
We introduce an adaptive output-sensitive Metropolis-Hastings algorithm for probabilistic models expressed as programs, Adaptive Lightweight Metropolis-Hastings (AdLMH). The algorithm extends Lightweight Metropolis-Hastings (LMH) by adjusting the probabilities of proposing random variables for modification to improve convergence of the program output. We show that AdLMH converges to the correct equilibrium distribution and compare convergence of AdLMH to that of LMH on several test problems to highlight different aspects of the adaptation scheme. We observe consistent improvement in convergence on the test problems.
Cite
@article{arxiv.1501.05677,
title = {Output-Sensitive Adaptive Metropolis-Hastings for Probabilistic Programs},
author = {David Tolpin and Jan Willem van de Meent and Brooks Paige and Frank Wood},
journal= {arXiv preprint arXiv:1501.05677},
year = {2015}
}