On the Equivalence Between Stochastic Tournament and Power-Law Ranking Selection and How to Implement Them Efficiently

Download PDF
Open Webpage

Duc-Cuong Dang, Andre Opris, Dirk Sudholt

Published: 2024, Last Modified: 07 May 2026PPSN (3) 2024EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Tournament selection is a popular parent selection mechanism in evolutionary algorithms. Bian and Qian (PPSN 2022) proved that choosing the tournament size uniformly at random, called stochastic tournament selection, in combination with crossover significantly improves the performance of NSGA-II on some benchmark functions. We show that this selection mechanism is asymptotically equivalent to the power-law ranking selection proposed in Covantes Osuna et al. (Theor. Comput. Sci. 832, 2020) with the exponent of 2. Thus asymptotic runtime bounds proven for one operator also hold when one operator is replaced with the other.