Uniform as Glass: Gliding over the Pareto Front with Neural Adaptive Preferences
Primary Area: optimization
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Keywords: multiobjective optimization; uniform design; solution modelling
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
TL;DR: Using a dynamic weight adjustment method within the MOEA/D framework to obtain uniformly distributed Pareto objectives
Abstract: Multiobjective optimization (MOO) is prevalent in numerous real-world applications, in which a Pareto front (PF) is constructed to display optima under various preferences. Previous methods commonly utilize the set of Pareto objectives (particles) to represent the entire Pareto front. However, the corresponding discrete distribution of the points on the PF is less studied, which may impede the generation of diverse and representative Pareto objectives in previous methods. To bridge the gap, we highlight in this paper the benefits of uniformly distributed Pareto objectives on the PF, which alleviate the limited diversity found in previous multiobjective optimization (MOO) approaches. In particular, we introduce new techniques for measuring and analyzing the uniformity of Pareto objectives, and accordingly propose a new method to generate asymptotically uniform Pareto objectives in an adaptive manner. Our proposed method is validated through experiments on real-world and synthetic problems, which demonstrates its efficacy in generating high-quality uniform Pareto objectives on the Pareto front.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 3545
Loading