Some Remarks on Replicated Simulated Annealing

TL;DR

This paper analyzes the replicated simulated annealing algorithm for training discrete neural network weights, providing convergence criteria and experimental validation on synthetic and real data.

Contribution

It offers explicit convergence criteria and insights into when the algorithm effectively samples from configurations, advancing understanding of its robustness.

Findings

Provides explicit convergence criteria for the algorithm.

Identifies conditions for successful sampling from configurations.

Validates the method through experiments on synthetic and real data.

Abstract

Recently authors have introduced the idea of training discrete weights neural networks using a mix between classical simulated annealing and a replica ansatz known from the statistical physics literature. Among other points, they claim their method is able to find robust configurations. In this paper, we analyze this so-called "replicated simulated annealing" algorithm. In particular, we explicit criteria to guarantee its convergence, and study when it successfully samples from configurations. We also perform experiments using synthetic and real data bases.