®®®® SIIA Público

Título del libro: Gecco 2023 Companion - Proceedings Of The 2023 Genetic And Evolutionary Computation Conference Companion
Título del capítulo: On the Effect of Temporal Heterogeneity on Selection Pressure of Evolutionary Algorithms

Autores UNAM:
CARLOS GERSHENSON GARCIA; CARLOS IGNACIO HERNANDEZ CASTELLANOS;
Autores externos:

Idioma:

Año de publicación:
2023
Palabras clave:

Multiobjective optimization; Parameter estimation; Deterministics; Evolutionary strategies; Exploration and exploitation; Multi-Objective Evolutionary Algorithm; Parameter control; Parameter tunning; Search spaces; Selection pressures; State of the art; Temporal heterogeneities; Evolutionary algorithms


Resumen:

State-of-the-art multi-objective evolutionary algorithms (MOEAs) are highly elitist. These algorithms usually employ a deterministic (µ+?) selection, where only the best individuals are selected for the next generation. However, for certain problems, it could be useful to have greater diversity to explore the search space. In this work, we aim to study the impact of temporal heterogeneity on the search to find a balance between exploration and exploitation by adding an extra parameter to evolutionary algorithms. This parameter controls the selection pressure going from (µ+?) to (µ,?) strategies. We analyzed classical evolutionary algorithms for single and multi-objective optimization (three and five objectives) and tested them on academic benchmarks. Our preliminary results indicate that there is a significant statistical difference at least for single-objective optimization problems. These results indicate that we should adapt the selection pressure for different problems and at different stages of the search. © 2023 Copyright held by the owner/author(s).


Entidades citadas de la UNAM: