A Gibbs distribution that learns from GA dynamics

TL;DR

This paper introduces a method to analyze genetic algorithm dynamics by learning Gibbs distributions from gene configurations, providing insights into their thermodynamic properties and performance.

Contribution

It presents a novel learning algorithm for Gibbs distributions based on Kullback-Leibler minimization, applied to models with complex energy landscapes.

Findings

Effective temperature scheduling behavior analyzed

Residual energy trends observed in simulations

Gibbs distribution learning captures GA statistical properties

Abstract

A general procedure of average-case performance evaluation for population dynamics such as genetic algorithms (GAs) is proposed and its validity is numerically examined. We introduce a learning algorithm of Gibbs distributions from training sets which are gene configurations (strings) generated by GA in order to figure out the statistical properties of GA from the view point of thermodynamics. The learning algorithm is constructed by means of minimization of the Kullback-Leibler information between a parametric Gibbs distribution and the empirical distribution of gene configurations. The formulation is applied to the solvable probabilistic models having multi-valley energy landscapes, namely, the spin glass chain and the Sherrington-Kirkpatrick model. By using computer simulations, we discuss the asymptotic behaviour of the effective temperature scheduling and the residual energy induced by the GA dynamics.