Integration of Knowledge Discovery into MOEA/D - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Integration of Knowledge Discovery into MOEA/D

Résumé

Extracting knowledge from solutions and then using it to guide the search is a complex task, which has not been highly explored in a discrete multi-objective optimization context. Considering the papers on that subject leads to the following terminology for Knowledge Discovery (KD) processes. A KD process is built upon two main procedures called Knowledge Extraction (K ext) and Knowledge Injection (K inj). The K ext procedure aims to extract problem-related knowledge from one or several solutions. Then the extracted knowledge can be used by the K inj procedure to build new solutions taking into account past iterations. In this article, we investigate how a KD mechanism can be integrated into MOEA/D. To that purpose, we consider a bi-objective Vehicle Routing Problem with Time Windows (bVRPTW). In this problem, we minimize both the total traveling time and the total waiting time of drivers. With these two objectives, we obtain more diverse and dense fronts than those obtained when minimizing the number of vehicles and the total traveling time. A study of existing works in KD and its hybridization with metaheuristics [2] leads to four main questions : What/Where/When/How is the knowledge extracted/injected ? Question What is problem-dependent, since each problem may have specific relevant knowledge. In the context of this article, we use sequences of consecutive customers, excluding the depot, from generated solutions. Questions Where and When are algorithm-dependent since the extraction and injection steps have to be integrated into the process of the algorithm. Question How deals with strategies that are used during the KD process (e.g. intensification or diversification). The answer to this question needs to take into account the multi-objective context of the problem. Our contribution focuses on this question and is detailed in Section 2.
Fichier principal
Vignette du fichier
ROADEF_2023-1.pdf (257.98 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04206237 , version 1 (13-09-2023)

Identifiants

  • HAL Id : hal-04206237 , version 1

Citer

Clément Legrand, Diego Cattaruzza, Marie-Eléonore Kessaci, Laetitia Jourdan. Integration of Knowledge Discovery into MOEA/D. ROADEF 2023, Feb 2023, Rennes, France. ⟨hal-04206237⟩
21 Consultations
5 Téléchargements

Partager

Gmail Facebook X LinkedIn More