Decomposition of large nonnegative tensors using memetic algorithms with application to environmental data analysis - Aix-Marseille Université Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2019

Decomposition of large nonnegative tensors using memetic algorithms with application to environmental data analysis

Résumé

In this article, we address the problem of the Canonical Polyadic decomposition (or Candecomp/Parafac Decomposition) of large N-way tensors under nonnegativity constraints. This problem is usually carried out using algebraic or iterative (alternating or all at once) deterministic optimization algorithms. Here, we focus on stochastic approaches and more precisely on memetic algorithms. Different variants of these algorithms are suggested and compared. A partial cost function and other optimization tools are introduced to reduce the complexity of the problem at hand. Different (either deterministic or stochastic) strategies concerning the local search step are also considered. This leads to new algorithms which are analysed thanks to computer simulations and compared with state of the art algorithms. When the tensor rank is unknown, we also propose a solution to estimate it. Finally, our approach is tested on a real experimental water monitoring application where the tensor rank is unknown.
Fichier principal
Vignette du fichier
MEMETIC_Article_submissionVersionEnvoyée.pdf (761.85 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02389879 , version 1 (02-12-2019)

Identifiants

  • HAL Id : hal-02389879 , version 1

Citer

Sylvain Maire, Xuan Vu, Caroline Chaux, Cyril Prissette, Nadège Thirion-Moreau. Decomposition of large nonnegative tensors using memetic algorithms with application to environmental data analysis. 2019. ⟨hal-02389879⟩
196 Consultations
132 Téléchargements

Partager

Gmail Facebook X LinkedIn More