English

Moving Target Monte Carlo

Computation 2020-03-11 v1 Data Structures and Algorithms Optimization and Control Probability Machine Learning

Abstract

The Markov Chain Monte Carlo (MCMC) methods are popular when considering sampling from a high-dimensional random variable x\mathbf{x} with possibly unnormalised probability density pp and observed data d\mathbf{d}. However, MCMC requires evaluating the posterior distribution p(xd)p(\mathbf{x}|\mathbf{d}) of the proposed candidate x\mathbf{x} at each iteration when constructing the acceptance rate. This is costly when such evaluations are intractable. In this paper, we introduce a new non-Markovian sampling algorithm called Moving Target Monte Carlo (MTMC). The acceptance rate at nn-th iteration is constructed using an iteratively updated approximation of the posterior distribution an(x)a_n(\mathbf{x}) instead of p(xd)p(\mathbf{x}|\mathbf{d}). The true value of the posterior p(xd)p(\mathbf{x}|\mathbf{d}) is only calculated if the candidate x\mathbf{x} is accepted. The approximation ana_n utilises these evaluations and converges to pp as nn \rightarrow \infty. A proof of convergence and estimation of convergence rate in different situations are given.

Keywords

Cite

@article{arxiv.2003.04873,
  title  = {Moving Target Monte Carlo},
  author = {Haoyun Ying and Keheng Mao and Klaus Mosegaard},
  journal= {arXiv preprint arXiv:2003.04873},
  year   = {2020}
}
R2 v1 2026-06-23T14:10:32.116Z