LibMOON: A Gradient-based MultiObjective OptimizatioN Library in PyTorch

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Xiaoyuan, Zhao, Liang, Yu, Yingying, Lin, Xi, Chen, Yifan, Zhao, Han, Zhang, Qingfu
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917800360017920
author Zhang, Xiaoyuan
Zhao, Liang
Yu, Yingying
Lin, Xi
Chen, Yifan
Zhao, Han
Zhang, Qingfu
author_facet Zhang, Xiaoyuan
Zhao, Liang
Yu, Yingying
Lin, Xi
Chen, Yifan
Zhao, Han
Zhang, Qingfu
contents Multiobjective optimization problems (MOPs) are prevalent in machine learning, with applications in multi-task learning, learning under fairness or robustness constraints, etc. Instead of reducing multiple objective functions into a scalar objective, MOPs aim to optimize for the so-called Pareto optimality or Pareto set learning, which involves optimizing more than one objective function simultaneously, over models with thousands / millions of parameters. Existing benchmark libraries for MOPs mainly focus on evolutionary algorithms, most of which are zeroth-order / meta-heuristic methods that do not effectively utilize higher-order information from objectives and cannot scale to large-scale models with thousands / millions of parameters. In light of the above gap, this paper introduces LibMOON, the first multiobjective optimization library that supports state-of-the-art gradient-based methods, provides a fair benchmark, and is open-sourced for the community.
format Preprint
id arxiv_https___arxiv_org_abs_2409_02969
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle LibMOON: A Gradient-based MultiObjective OptimizatioN Library in PyTorch
Zhang, Xiaoyuan
Zhao, Liang
Yu, Yingying
Lin, Xi
Chen, Yifan
Zhao, Han
Zhang, Qingfu
Mathematical Software
Machine Learning
Optimization and Control
Multiobjective optimization problems (MOPs) are prevalent in machine learning, with applications in multi-task learning, learning under fairness or robustness constraints, etc. Instead of reducing multiple objective functions into a scalar objective, MOPs aim to optimize for the so-called Pareto optimality or Pareto set learning, which involves optimizing more than one objective function simultaneously, over models with thousands / millions of parameters. Existing benchmark libraries for MOPs mainly focus on evolutionary algorithms, most of which are zeroth-order / meta-heuristic methods that do not effectively utilize higher-order information from objectives and cannot scale to large-scale models with thousands / millions of parameters. In light of the above gap, this paper introduces LibMOON, the first multiobjective optimization library that supports state-of-the-art gradient-based methods, provides a fair benchmark, and is open-sourced for the community.
title LibMOON: A Gradient-based MultiObjective OptimizatioN Library in PyTorch
topic Mathematical Software
Machine Learning
Optimization and Control
url https://arxiv.org/abs/2409.02969