Estimating Ratios of Normalizing Constants Using Linked Importance Sampling

TL;DR

This paper introduces Linked Importance Sampling (LIS), a novel method for estimating ratios of normalizing constants that can outperform Annealed Importance Sampling (AIS) in accuracy for certain problems, with broad applicability.

Contribution

The paper develops LIS, a new importance sampling technique that improves ratio estimation accuracy over AIS by using bridge sampling-like estimates in an extended state space.

Findings

LIS can be more accurate than AIS for some problems.

LIS provides unbiased estimates even without full Markov chain equilibrium.

Performance varies depending on problem characteristics.

Abstract

Ratios of normalizing constants for two distributions are needed in both Bayesian statistics, where they are used to compare models, and in statistical physics, where they correspond to differences in free energy. Two approaches have long been used to estimate ratios of normalizing constants. The `simple importance sampling' (SIS) or `free energy perturbation' method uses a sample drawn from just one of the two distributions. The `bridge sampling' or `acceptance ratio' estimate can be viewed as the ratio of two SIS estimates involving a bridge distribution. For both methods, difficult problems must be handled by introducing a sequence of intermediate distributions linking the two distributions of interest, with the final ratio of normalizing constants being estimated by the product of estimates of ratios for adjacent distributions in this sequence. Recently, work by Jarzynski, and independently by Neal, has shown how one can view such a product of estimates, each based on simple importance sampling using a single point, as an SIS estimate on an extended state space. This `Annealed Importance Sampling' (AIS) method produces an exactly unbiased estimate for the ratio of normalizing constants even when the Markov transitions used do not reach equilibrium. In this paper, I show how a corresponding `Linked Importance Sampling' (LIS) method can be constructed in which the estimates for individual ratios are similar to bridge sampling estimates. I show empirically that for some problems, LIS estimates are much more accurate than AIS estimates found using the same computation time, although for other problems the two methods have similar performance. Linked sampling methods similar to LIS are useful for other purposes as well.