Complex Systems

A Note on Boltzmann Tournament Selection for Genetic Algorithms and Population-Oriented Simulated Annealing Download PDF

David E. Goldberg
University of Illinois at Urbana--Champaign, Urbana IL 61801 USA

Abstract

This note describes a selection procedure for genetic algorithms called Boltzmann tournament selection. As simulated annealing evolves a Boltzmann distribution in time using the Metropolis algorithm or a logistic acceptance mechanism, Boltzmann tournament selection evolves a Boltzmann distribution across a population and time using pairwise probabilistic acceptance and anti-acceptance mechanisms. In this note, the motivation, the theory of operation, some proof-of-principle computational experiments, and a Pascal implementation of the algorithm are presented. The efficient use of Boltzmann tournament selection on parallel hardware and its connection to other niching mechanisms are also considered.