Approximate Survey Propagation for Statistical Inference

TL;DR

This paper introduces Approximate Survey Propagation (ASP), an algorithm that accounts for glassy systems in low-rank matrix estimation, matching the 1RSB physics equations and outperforming AMP in model mismatch scenarios.

Contribution

The paper derives ASP from 1RSB equations, providing a concrete algorithmic interpretation and analyzing its performance in low-rank matrix estimation.

Findings

ASP converges in larger regimes than AMP under model mismatch.

ASP can achieve near Bayes-optimal error when parameters are tuned to restore Nishimori conditions.

ASP reproduces the 1RSB fixed-point equations, linking physics and algorithmic performance.

Abstract

Approximate message passing algorithm enjoyed considerable attention in the last decade. In this paper we introduce a variant of the AMP algorithm that takes into account glassy nature of the system under consideration. We coin this algorithm as the approximate survey propagation (ASP) and derive it for a class of low-rank matrix estimation problems. We derive the state evolution for the ASP algorithm and prove that it reproduces the one-step replica symmetry breaking (1RSB) fixed-point equations, well-known in physics of disordered systems. Our derivation thus gives a concrete algorithmic meaning to the 1RSB equations that is of independent interest. We characterize the performance of ASP in terms of convergence and mean-squared error as a function of the free Parisi parameter s. We conclude that when there is a model mismatch between the true generative model and the inference model, the performance of AMP rapidly degrades both in terms of MSE and of convergence, while ASP converges in a larger regime and can reach lower errors. Among other results, our analysis leads us to a striking hypothesis that whenever s (or other parameters) can be set in such a way that the Nishimori condition $M=Q>0$ is restored, then the corresponding algorithm is able to reach mean-squared error as low as the Bayes-optimal error obtained when the model and its parameters are known and exactly matched in the inference procedure.