EURO 2025 Leeds
Abstract Submission

1851. MODA - A new C++/Python library with algorithms and data structures for multiobjective optimization

Invited abstract in session WA-51: Advances in multiobjective optimization software, stream Multiobjective and vector optimization.

Wednesday, 8:30-10:00
Room: Parkinson B22

Authors (first author is the speaker)

1. Andrzej Jaszkiewicz
Faculty of Computing, Poznan University of Technology
2. Jakub Dutkiewicz
Poznan University of Technology
3. Piotr Zielniewicz
Institute of Computing Science, Poznan University of Technology

Abstract

Advancements in (evolutionary) multiobjective optimization can be achieved not only by developing new MO methods but also by enhancing the efficiency and/or accuracy of fundamental algorithmic procedures used across different MO techniques. In recent years, we have introduced various data structures and algorithms to address tasks such as:
- Efficiently updating the Pareto archive, including verifying whether a new solution is nondominated, adding it to the archive if so, and removing dominated solutions, utilizing the ND-Tree data structure.
- Identifying the solution that minimizes or maximizes a (modified) Chebyshev scalarizing function over a finite solution set, also leveraging ND-Tree.
- Efficiently estimating hypervolume or R2 (contribution) with the ND-Tree.
- Computing hypervolume using the improved quick hypervolume algorithm.
- Determining guaranteed bounds for hypervolume (contribution).
- Enhancing the accuracy of Monte Carlo-based hypervolume (contribution) estimation.
- Identifying the extreme (minimum or maximum) hypervolume contribution/contributor within a solution set.
- Hypervolume subset selection using a lazy incremental or decremental greedy approach.
- Performing on-the-fly updates of hypervolume values.
- Exactly calculating the R2 quality indicator.
We present a new MO library called MODA (MultiObjective Data structures and Algorithms) in C++ with Python interface that implements these algorithms and data structures.

Keywords

Status: accepted

Back to the list of papers