Multi-objective Learning Using HV Maximization

Multi-objective Learning Using HV Maximization

Deist, Timo M. (Centrum Wiskunde & Informatica (CWI))
Grewal, M. (Centrum Wiskunde & Informatica (CWI))
Dankers, Frank J.W.M. (Leiden University Medical Center)
Alderliesten, T. (Leiden University Medical Center)
Bosman, P.A.N. (TU Delft Algorithmics; Centrum Wiskunde & Informatica (CWI))

Emmerich, Michael (editor)
Deutz, André (editor)
Wang, Hao (editor)
Kononova, Anna V. (editor)
Naujoks, Boris (editor)
Li, Ke (editor)
Miettinen, Kaisa (editor)
Yevseyeva, Iryna (editor)

2023

Real-world problems are often multi-objective, with decision-makers unable to specify a priori which trade-off between the conflicting objectives is preferable. Intuitively, building machine learning solutions in such cases would entail providing multiple predictions that span and uniformly cover the Pareto front of all optimal trade-off solutions. We propose a novel approach for multi-objective training of neural networks to approximate the Pareto front during inference. In our approach, we train the neural networks multi-objectively using a dynamic loss function, wherein each network’s losses (corresponding to multiple objectives) are weighted by their hypervolume maximizing gradients. Experiments on different multi-objective problems show that our approach returns well-spread outputs across different trade-offs on the approximated Pareto front without requiring the trade-off vectors to be specified a priori. Further, results of comparisons with the state-of-the-art approaches highlight the added value of our proposed approach, especially in cases where the Pareto front is asymmetric.

Hypervolume
Multi-objective learning
Multi-objective optimization
Neural networks
Pareto front

http://resolver.tudelft.nl/uuid:593daea0-ccb1-417d-99b7-25542168d266

https://doi.org/10.1007/978-3-031-27250-9_8

Springer, Cham

2023-09-09

978-3-031-27249-3

Evolutionary Multi-Criterion Optimization - 12th International Conference, EMO 2023, Proceedings

12th International Conference on Evolutionary Multi-Criterion Optimization, EMO 2023, 2023-03-20 → 2023-03-24, Leiden, Netherlands

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 0302-9743, 13970

Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.

Institutional Repository

conference paper

© 2023 Timo M. Deist, M. Grewal, Frank J.W.M. Dankers, T. Alderliesten, P.A.N. Bosman