The influence of fitness caching on modern evolutionary methods and fair computation load measurement

Michal W. Przewozniczek, Marcin M. Komarnicki · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2018

Any evolutionary method may store the fitness values for the genotypes it has already rated. Any time the fitness is to be computed, the check may be made if the fitness for the same genotype was not computed earlier. If so, then instead of re-evaluating the same genotype, the stored value from the repository may be returned. Such technique will be denoted as fitness caching. It is easy to implement in any evolutionary method, and it minimizes the number of fitness function evaluations (FFE), which is desirable. Despite its simplicity fitness caching may significantly affect the computation load spent on fitness computation. Moreover, it may cause that the FFE will not be a reliable computation load measure.

Read the paper · More papers on PaperTik