Application des copules à l'estimation de fronts de Pareto

Abstract : Optimization studies generally start by randomly sampling in the variable space to provide an initial population or to create a surrogate model. In particular, in the multi-objective case, the result is a set of non-dominated points which provides little information on the true Pareto front. We propose to study this problem from the point of view of multivariate analysis, introducing a probabilistic framework with the use of copulas. Specifically, the Pareto front appears as a zero level line of the multivariate distribution of the samples in the objective space. In particular, using Archimedean copulas provides analytical expression for estimation of Pareto fronts. The corresponding estimation procedure is described and illustrated on several examples.
Document type :
Conference papers
Complete list of metadatas

https://hal-emse.ccsd.cnrs.fr/emse-01152513
Contributor : Florent Breuil <>
Submitted on : Monday, May 18, 2015 - 9:47:06 AM
Last modification on : Wednesday, May 8, 2019 - 1:24:52 AM

Identifiers

  • HAL Id : emse-01152513, version 1

Citation

Mickaël Binois, Didier Rullière, Olivier Roustant. Application des copules à l'estimation de fronts de Pareto. 47èmes Journées de Statistique de la SFdS, Société Française de Statistique, Jun 2015, Lille, France. ⟨emse-01152513⟩

Share

Metrics

Record views

204