Aligned Multi-Objective Optimization

Yonathan Efroni, Daniel Jiang, Ben Kretzu, Jalaj Bhandari, Zheqing Zhu, Karen Ullrich

Keywords: multi task learning, multi objective optimization

TL;DR: We investigate a multi-objective setting where different objectives have a shared optimal solution and develop a gradient descent algorithm with provably improved guarantees.

Abstract: To date, the multi-objective optimization literature has mainly focused on conflicting objectives, studying the Pareto front or requiring users to balance trade-offs. Yet, in machine learning practice, there are many scenarios where such conflict does not take place. Recent findings from multi-task learning, reinforcement learning, and LLMs show that diverse related tasks can enhance performance across objectives simultaneously. Despite this evidence, such phenomenon has not been examined from an optimization perspective. This leads to a lack of generic gradient-based methods that can scale to scenarios with a large number of related objectives. To address this gap, we introduce the Aligned Multi-Objective Optimization framework, propose the \texttt{AMOOO} algorithm, and provide theoretical guarantees of its superior performance compared to naive approaches.

Submission Number: 37