Parameter-less, population-sizing DSMGA-II
Marcin M. Komarnicki, Michal W. Przewozniczek · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2019
Limiting the number of required settings is an important part of any evolutionary method development. The final objective of this process is a method version that is parameter-less. Based on the research results presented that far, the leading methods in the combinatorial optimization are Linkage Tree Genetic Algorithm (LTGA), Parameter-less Population Pyramid (P3) and Dependency Structure Matrix Genetic Algorithm II (DSMGA-II). P3 was originally proposed as a parameter-less method, while LTGA and DSMGA-II in their original propositions both require one parameter that is the population size. Recently a population-sizing technique was used to propose a parameter-less version of LTGA (psLTGA). However, the population-sizing was not introduced in DSMGA-II that misses its effective parameter-less version. Therefore, to fill this gap, in this paper we propose a Population-sizing DSMGA-II (psDSMGA-II) that is parameter-less. We also show that psDSMGA-II is more effective than its predecessor and that it may successfully compete with psLTGA and P3.