Evolutionary Algorithms with On-the-Fly Population Size Adjustment

Abstract

Abstract. In this paper we evaluate on-the-fly population (re)sizing mechanisms for evolutionary algorithms (EAs). Evaluation is done by an experimental comparison, where the contestants are various existing methods and a new mechanism, introduced here. These comparisons consider EA performance in terms of success rate, speed, and solution quality, measured on a variety of fitness landscapes. These landscapes are created by a generator that allows for gradual tuning of their characteristics. Our test suite covers a wide span of landscapes ranging from a smooth one-peak landscape to a rugged 1000-peak one. The experiments show that the population (re)sizing mechanisms exhibit significant differences in speed, measured by the number of fitness evaluations to a solution and the best EAs with adaptive population resizing outperform the traditional genetic algorithm (GA) by a large margin.

Variable population size and evolution acceleration: a case study with a parallel evolutionary algorithm

“…These statistics are depicted as bar charts with error intervals in Figs. 6,7,8,9. From these figures we conclude that the rate of best fitness improvement is Best fitness improvement Micro-step Recovery duration D = 2,000 ms Fig.…”

“…Later, Eiben et al [7] suggest to use the pace of fitness improvements as the signal to adjust population size dynamically in the Population Resizing on Fitness Improvement GA (PRoFIGA). Similarly in genetic programming systems, Tomassini et al [28] implement a dynamic population size algorithm using fitness progression as the signal to delete over-sized and worse-fit individuals or to insert mutated best-fit individuals under certain conditions.…”

Incremental Particle Swarm-Guided Local Search for Continuous Optimization

“…[2,19,4,13,14,9]). An exception to this approach is the work of Auger and Hansen [3] in which the population size of a CMA-ES algorithm is doubled each time it is restarted.…”

Millipede, an Extended Representation for Genetic Algorithms

“…Various other attempts at improving GA representation have been proposed in the past, including gray coding [2], adjusting population size on-the-fly [3], [4], the grouping genetic algorithm encoding structure [5], and the proportional GA. Banzhaf's GPM [6] looked at many genotypes mapping into one phenotype. In this paper we look at one genotype mapping into many phenotypes instead.…”