New Computational Models for the Choquet Integral - Archive ouverte HAL Access content directly
Conference Papers Year :

New Computational Models for the Choquet Integral

Hugo Martin
  • Function : Author
  • PersonId : 734824
  • IdHAL : hugomartin
Patrice Perny

Abstract

Multiobjective optimization is a central problem in a wide range of contexts, such as multi-agent optimization and multicriteria decision support or decision under risk and uncertainty. The presence of several objectives leads to multiple non-dominated solutions and requires the use of a sophisticated decision model allowing various attitudes towards preference aggregation. The Choquet Integral is one of the most expressive parameterized models introduced in decision theory to scalarize performance vectors and support decision making. However, its use in optimization contexts raises computational issues. This paper proposes new computational models based on mathematical programming to optimize the Choquet integral on implicit sets. A new linearization of the Choquet integral exploiting the vertices of the core of the convex capacity is proposed, combined with a constraint generation algorithm. Then the computational model is extended to the bipolar Choquet Integral to allow asymmetric aggregation with respect to a specific reference point.
No file

Dates and versions

hal-02865253 , version 1 (11-06-2020)

Identifiers

  • HAL Id : hal-02865253 , version 1

Cite

Hugo Martin, Patrice Perny. New Computational Models for the Choquet Integral. 24th European Conference on Artificial Intelligence - ECAI 2020, Aug 2020, Santiago, Spain. ⟨hal-02865253⟩
62 View
0 Download

Share

Gmail Facebook Twitter LinkedIn More