EUROPT 2024
Abstract Submission

202. Solving vector optimization problems with ADMM

Invited abstract in session WF-4: Multiobjective Optimization I, stream Multiobjective optimization.

Wednesday, 16:20 - 18:00
Room: M:M

Authors (first author is the speaker)

1. Daniel Hernandez Escobar
Department of Information Technology, Uppsala University
2. Joakim da Silva
Elekta
3. Jens Sjölund
Department of Information Technology, Uppsala University

Abstract

We consider the numerical solution to vector optimization problems. We focus mainly on convex problems whose preference order is defined by a generalized inequality. To approximate the set of efficient solutions, we employ the Alternating Direction Method of Multipliers and a parallel strategy. Although our approach may produce duplicate solutions, it can leverage GPUs or TPUs to achieve fast convergence. We illustrate this by solving multi-objective linear programs. Moreover, we outline how to adapt this technique to solve other problem classes, for instance, when a Lorentz cone defines the generalized inequality.

Keywords

Status: accepted


Back to the list of papers