Aus Aifbportal
Wechseln zu:Navigation, Suche

Incorporating user preferences in multi-objective evolutionary algorithms

Marlon Alexander Braun

Informationen zur Arbeit

Abschlussarbeitstyp: Bachelor
Betreuer: Pradyumn Kumar ShuklaHartmut Schmeck
Forschungsgruppe: Effiziente Algorithmen

Archivierungsnummer: 3435
Abschlussarbeitsstatus: Abgeschlossen
Beginn: unbekannt
Abgabe: 11. Juli 2011

Weitere Informationen

For many years evolutionary algorithms have shown to be capable of solving multi-objective optimization problems. Nearly all of these algorithms aim at finding an approximation of the Pareto optimal front, the set of solutions that are not dominated by any other selection of decision variables. Considering user preferences, however, not all Pareto optimal solu- tions necessarily present an interesting choice. The concept of proper Pareto optimality tries to bound the trade-off between objective values. The notion has been successfully im- plemented in deterministic search methods, but has so far only received little attention in evolutionary optimization. This thesis proposes modifications of existing state-of-the-art algorithms that guide the search towards proper Pareto optimal solutions. An extensive set of different benchmark problems is used to assess the performance of these algorithms. Computational results indicate that the modified algorithms are able to find the complete solution set of preferred regions.