Analysis and optimization of weighted ensemble sampling

TL;DR

This paper presents a mathematical framework for weighted ensemble sampling, proving its unbiasedness and demonstrating how coarse models can optimize replica allocation for efficient probability estimation in molecular dynamics.

Contribution

It introduces a general unbiasedness proof for WE sampling, including adaptive binning, and shows how coarse models can optimize replica distribution.

Findings

WE sampling is unbiased in a broad setting.

Coarse models can optimize replica allocation.

Enhanced efficiency in probability calculations for molecular dynamics.

Abstract

We give a mathematical framework for weighted ensemble (WE) sampling, a binning and resampling technique for efficiently computing probabilities in molecular dynamics. We prove that WE sampling is unbiased in a very general setting that includes adaptive binning. We show that when WE is used for stationary calculations in tandem with a coarse model, the coarse model can be used to optimize the allocation of replicas in the bins.