Generalizing self-normalized importance sampling with couplings

TL;DR

This paper introduces a novel framework for importance sampling that leverages couplings and joint proposals to reduce variance and improve estimation accuracy in challenging statistical problems.

Contribution

It proposes a new formulation of importance sampling using joint proposals and couplings, enabling better adaptation and variance control compared to existing methods.

Findings

Improved estimation accuracy in rare event scenarios.

Enhanced adaptability of importance sampling via joint proposals.

Successful application to Bayesian prediction with misspecified models.

Abstract

An essential problem in statistics and machine learning is the estimation of expectations involving PDFs with intractable normalizing constants. The self-normalized importance sampling (SNIS) estimator, which normalizes the IS weights, has become the standard approach due to its simplicity. However, the SNIS has been shown to exhibit high variance in challenging estimation problems, e.g, involving rare events or posterior predictive distributions in Bayesian statistics. Further, most of the state-of-the-art adaptive importance sampling (AIS) methods adapt the proposal as if the weights had not been normalized. In this paper, we propose a framework that considers the original task as estimation of a ratio of two integrals. In our new formulation, we obtain samples from a joint proposal distribution in an extended space, with two of its marginals playing the role of proposals used to estimate each integral. Importantly, the framework allows us to induce and control a dependency between both estimators. We propose a construction of the joint proposal that decomposes in two (multivariate) marginals and a coupling. This leads to a two-stage framework suitable to be integrated with existing or new AIS and/or variational inference (VI) algorithms. The marginals are adapted in the first stage, while the coupling can be chosen and adapted in the second stage. We show in several examples the benefits of the proposed methodology, including an application to Bayesian prediction with misspecified models.