Offline Multi-Task Multi-Objective Data-Driven Evolutionary Algorithm with Language Surrogate Model and Implicit Q-Learning

TL;DR

This paper introduces Q-MetaSur, a novel language model-based surrogate for multi-task multi-objective offline optimization, improving approximation accuracy and optimization performance in complex problems.

Contribution

Proposes a language model-based surrogate with a two-stage training strategy for multi-task multi-objective offline optimization, addressing limitations of existing surrogate models.

Findings

Q-MetaSur outperforms baseline models in objective approximation accuracy.

Q-MetaSur enhances evolutionary algorithm convergence and Pareto optimality.

Empirical results on CEC2019 benchmark validate effectiveness.

Abstract

Data-driven evolutionary algorithms has shown surprising results in addressing expensive optimization problems through robust surrogate modeling. Though promising, existing surrogate modeling schemes may encounter limitations in complex optimization problems with many sub-objectives, which rely on repeated and tedious approximation. To address such technical gap, we propose Q-MetaSur as a plug-and-play surrogate modeling scheme capable of providing unified and generalized surrogate learning. Specifically, we consider multi-task-multi-objective optimization~(MTMOO) in offline setting. Several key designs are proposed: 1) we transform objective approximation into sequence-to-sequence modeling where MTMOO problem can be represented by tenxual tokenization. To operate under such auto-regressive modeling, we introduce a Large Language Model-based surrogate model that first encodes a MTMOO instance and then decodes objective values of unseen decision variables. To ensure stability in training the proposed model, we propose a two-stage offline training strategy that operates as a synergy of supervised tuning and RL fine-tuning, which first exploits offline dataset to fit existing knowledge and then leverages RL to enhance model's generalization performance. Extensive empirical results on the CEC2019 benchmark demonstrate that Q-MetaSur not only outperforms representative surrogate baselines in objective approximation accuracy, but also helps underlying evolutionary algorithms achieve both desired optimization convergence and improved pareto optimality.