A characteristic of Bennett's acceptance ratio method

TL;DR

This paper proves that Bennett's acceptance ratio method has a convex mean square error, leading to an optimal ratio of forward and reverse samples for minimal error, and explores its relation to Jarzynski estimators.

Contribution

It establishes the convexity of the mean square error of Bennett's acceptance ratio method and identifies the unique optimal sampling ratio.

Findings

Convexity of the mean square error is proven.

An optimal ratio of forward to reverse samples minimizes error.

Insights into the relation with Jarzynski estimators are provided.

Abstract

A powerful and well-established tool for free-energy estimation is Bennett's acceptance ratio method. Central properties of this estimator, which employs samples of work values of a forward and its time reversed process, are known: for given sets of measured work values, it results in the best estimate of the free-energy difference in the large sample limit. Here we state and prove a further characteristic of the acceptance ratio method: the convexity of its mean square error. As a two-sided estimator, it depends on the ratio of the numbers of forward and reverse work values used. Convexity of its mean square error immediately implies that there exists an unique optimal ratio for which the error becomes minimal. Further, it yields insight into the relation of the acceptance ratio method and estimators based on the Jarzynski equation. As an application, we study the performance of a dynamic strategy of sampling forward and reverse work values.