Nouvelles heuristiques de voisinage et mémétiques pour le problème Maximum de Parcimonie

Abstract : Phylogenetic reconstruction aims at reconstructing the evolutionary history of a set of species, represented by a tree. Among the reconstruction methods, the Maximum Parsimony (MP) problem consists, given a set of aligned sequences to find a binary tree, whose leaves are associated to the sequences and which minimizes the parsimony score. Traditionally, existing resolution approaches of this NP-complete problem apply basic heuristic methods, like greedy algorithms and local search. One of the difficulties concerns the handling of binary trees and the definition of tree neighborhoods. In this thesis, we first focus on an improvement of descent algorithms. We empirically show the limits of the existing tree neighborhoods, and introduce a progressive neighborhood which evolves during the search to limit the evaluation of inappropriate neighbors. This algorithm is combined with a genetic algorithm which uses a specific tree crossover based on topological distances between each pair of leaves. This memetic algorithm shows very competitive results, both on real benchmarks taken from the literature as well as with randomly generated instances.
Document type :
Theses
Complete list of metadatas

Cited literature [167 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00256670
Contributor : Adrien Goëffon <>
Submitted on : Saturday, February 16, 2008 - 2:11:21 PM
Last modification on : Friday, May 10, 2019 - 12:50:14 PM
Long-term archiving on : Friday, September 28, 2012 - 9:55:34 AM

Identifiers

  • HAL Id : tel-00256670, version 1

Citation

Adrien Goëffon. Nouvelles heuristiques de voisinage et mémétiques pour le problème Maximum de Parcimonie. Autre [cs.OH]. Université d'Angers, 2006. Français. ⟨tel-00256670⟩

Share

Metrics

Record views

484

Files downloads

652