Skip to Main content Skip to Navigation
Theses

Extraire et exploiter la connaissance pour mieux optimiser

Lucien Mousin 1, 2, 3, 4
1 ORKAD - Operational Research, Knowledge And Data
CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille (CRIStAL) - UMR 9189
4 DOLPHIN - Parallel Cooperative Multi-criteria Optimization
Inria Lille - Nord Europe, CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille (CRIStAL) - UMR 9189
Abstract : Large-scale combinatorial optimization problems are generally hard to solve optimally due to expensive computation times. In order to tackle this problem, approximation algorithms such as heuristics and metaheuristics are used to quickly find approximate solutions. Heuristics are problem-specific approaches that provide solutions very quickly. Metaheuristics are generic approaches to finding solutions with good quality for several problems. Each of these approaches has its advantages and disadvantages. We propose in this thesis to use the advantages of these two approaches, that is to say, to integrate specific knowledge associated to a problem as heuristics do, within the mechanisms of metaheuristics in order to design new effective approaches. In this thesis report, we first review knowledge integration approaches in the literature to propose a taxonomy for a classification of these approaches. Then we focus on the integration of knowledge in two different problems: the No-Wait Flowshop Scheduling Problem, and the Feature Selection Problem in a classification context. Finally we study the impact of the integration of knowledge on these problems with different metaheuristic's mechanisms, that is initialization procedure, neighborhood operator and neighborhood selection.
Complete list of metadatas

Cited literature [109 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/tel-01989899
Contributor : Lucien Mousin <>
Submitted on : Tuesday, January 22, 2019 - 4:08:12 PM
Last modification on : Wednesday, May 13, 2020 - 5:02:08 PM
Document(s) archivé(s) le : Tuesday, April 23, 2019 - 5:11:06 PM

File

Thèse - Lucien Mousin.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-01989899, version 1

Collections

Citation

Lucien Mousin. Extraire et exploiter la connaissance pour mieux optimiser. Informatique [cs]. Université de Lille, 2018. Français. ⟨tel-01989899⟩

Share

Metrics

Record views

134

Files downloads

157