A Unified Framework for Gradient Aggregation in Multi-Objective Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hu, Zeou, Ho, Kelvin, Yu, Yaoliang
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916062327472128
author Hu, Zeou
Ho, Kelvin
Yu, Yaoliang
author_facet Hu, Zeou
Ho, Kelvin
Yu, Yaoliang
contents Many machine learning problems involve multiple inherent trade-offs that are best addressed by gradient-based multi-objective optimization (MOO) algorithms. Existing methods are often proposed with various motivations, analyzed case by case, and differ algorithmically in how the component gradients are aggregated at each step. In this work, we develop a unifying framework for gradient aggregation in MOO, establishing (optimal) rates of convergence to Pareto stationarity, the standard measure of performance in MOO. Central to our analysis is a sufficient alignment condition, from which we derive a theorem showing that non-conflicting directions, when chosen within the convex hull of gradients, form a fundamental sufficient condition for convergence. We further show that feasibility can be ensured through projection onto the dual cone, broadening the scope of methods that admit convergence guarantees. In parallel, we present a primal optimization perspective of gradient aggregation that encompasses established algorithms, clarifies their theoretical relationships, and enables the design of new variants. As an illustration, we introduce capped MGDA, derived from a CVaR-based formulation, and demonstrate its robustness in adversarial federated learning. Finally, we validate our theory through experiments on synthetic problems and practical benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2605_30452
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle A Unified Framework for Gradient Aggregation in Multi-Objective Optimization
Hu, Zeou
Ho, Kelvin
Yu, Yaoliang
Machine Learning
Artificial Intelligence
Optimization and Control
Many machine learning problems involve multiple inherent trade-offs that are best addressed by gradient-based multi-objective optimization (MOO) algorithms. Existing methods are often proposed with various motivations, analyzed case by case, and differ algorithmically in how the component gradients are aggregated at each step. In this work, we develop a unifying framework for gradient aggregation in MOO, establishing (optimal) rates of convergence to Pareto stationarity, the standard measure of performance in MOO. Central to our analysis is a sufficient alignment condition, from which we derive a theorem showing that non-conflicting directions, when chosen within the convex hull of gradients, form a fundamental sufficient condition for convergence. We further show that feasibility can be ensured through projection onto the dual cone, broadening the scope of methods that admit convergence guarantees. In parallel, we present a primal optimization perspective of gradient aggregation that encompasses established algorithms, clarifies their theoretical relationships, and enables the design of new variants. As an illustration, we introduce capped MGDA, derived from a CVaR-based formulation, and demonstrate its robustness in adversarial federated learning. Finally, we validate our theory through experiments on synthetic problems and practical benchmarks.
title A Unified Framework for Gradient Aggregation in Multi-Objective Optimization
topic Machine Learning
Artificial Intelligence
Optimization and Control
url https://arxiv.org/abs/2605.30452