An accelerated proximal gradient method for multiobjective optimization

Hiroki Tanabe, Ellen H. Fukuda, Nobuo Yamashita

Published 2022-02-22Version 1

Many descent methods for multiobjective optimization problems have been developed in recent years. In 2000, the steepest descent method was proposed for differentiable multiobjective optimization problems. Afterward, the proximal gradient method, which can solve composite problems, was also considered. However, the accelerated versions are not sufficiently studied. In this paper, we propose a multiobjective accelerated proximal gradient algorithm, in which we solve subproblems with terms that only appear in the multiobjective case. We also show the proposed method's global convergence rate ($O(1/k^2)$) under reasonable assumptions, using a merit function to measure the complexity. Moreover, we present an efficient way to solve the subproblem via its dual, and we confirm the validity of the proposed method through preliminary numerical experiments.