Generalized conditional gradient methods for multiobjective composite optimization problems with H{\"o}lder condition

TL;DR

This paper introduces generalized conditional gradient methods for multiobjective composite optimization, including a parameter-free approach that adapts step sizes without prior knowledge of problem-specific parameters, with demonstrated effectiveness on test problems.

Contribution

The paper proposes a novel parameter-free conditional gradient method for multiobjective composite optimization, overcoming limitations of parameter-dependent methods and enhancing practical applicability.

Findings

The parameter-free method achieves comparable convergence without prior parameter knowledge.

The methods effectively handle problems with indicator and uncertainty functions.

Convergence properties are rigorously established for the proposed algorithms.

Abstract

In this paper, we deal with multiobjective composite optimization problems, where each objective function is a combination of smooth and possibly non-smooth functions. We first propose a parameter-dependent conditional gradient method to solve this problem. The step size in this method requires prior knowledge of the parameters related to the H{\"o}lder continuity of the gradient of the smooth function. The convergence properties of this method are then established. Given that these parameters may be unknown or, if known, may not be unique, the first method may encounter implementation challenges or slow convergence. To address this, we further propose a parameter-free conditional gradient method that determines the step size using a local quadratic upper approximation and an adaptive line search strategy, eliminating the need for any problem-specific parameters. The performance of the proposed methods is demonstrated on several test problems involving the indicator function and an uncertainty function.