Short-Term Scientific Mission

Decision and Objective Space Representations (DO-REP)

Main theme: Single- and Multiobjective Optimisation
Grantee: Lara Loehken, Bergische Universität Wuppertal, Wuppertal, Germany
Host: Serpil Sayın, Koç University, Istanbul, Türkiye

Start date: 2025-07-14
End date: 2025-08-08
Awarded: 2025-04-04
Report approved: 2025-09-08

Description

Representation algorithms aim to provide a collection of discrete solutions to a multi-objective optimization problem. The solution set, called a representation, must ensure a certain quality with respect to a chosen measure or indicator. Recent research on representations mainly focuses on the objective space of multi- objective problems. In this STSM, we aim to investigate representations of solution sets in both the decision and objective space for continuous multi-objective problems. Our goal is to identify possible interrelations between representations of decision and objective space and ultimately seek ways to efficiently compute good representations with respect to suitable indicators for continuous multi-objective problems.

Achievements

During the STSM, a new research project on decision and objective space representations (DO-REP) in continuous multiobjective optimization was successfully initiated. Suitable representation indicators were investigated in decision and objective space and insights into interrelations between both spaces were obtained. Based on an existing approach to approximate Pareto optimal solutions, called Pareto tracing, different discretization strategies were examined to deliver high-quality representations. While no finalized algorithm exists yet, preliminary experiments show promising results. New benchmark test instances were designed to identify effects and interdependencies. The collaboration project will be continued and, once finalized, a research paper will be submitted.