Unifying All Species: LLM-based Hyper-Heuristics for Multi-objective Optimization
Keywords: Multi-Objective Optimization, Hierarchical Reflective Evolution, Ant Colony Optimization, Metaheuristic Algorithms, Combinatorial Optimization
TL;DR: This work presents the Multi-Objective Hierarchical Reflective Evolution (MHRE) framework, enhancing optimization through multi-objective approaches and unifying metaheuristic algorithms for superior performance in optimization problems.
Abstract: Optimization problems are fundamental across various fields, including logistics, machine learning, and bioinformatics, where challenges are often characterized by complexity, high dimensionality. Modeling the interplay among multiple objectives is beneficial for optimization. However, existing Neural Combinatorial Optimization (NCO) methods and Large Language Model (LLM)-based approaches show limitations in adaptability and computational efficiency, primarily focusing on single-objective optimization. In this paper, we propose a novel framework, Multi-Objective Hierarchical Reflective Evolution (MHRE), for optimizing and generating heuristics algorithms for a broad range of optimization problems. Specifically, we extend the optimization space of the conventional hyper-heuristic methodologies, which allows us to unify similarity algorithms. We successfully construct Generalized Evolutionary Metaheuristic Algorithm (GEMA) for unifying metaheuristic algorithms. Yielding improved performance in experimental results. To show the performance of our method, we further applied the MHRE framework to optimize the Ant Colony Optimization (ACO) algorithm, achieving state-of-the-art results on random TSP problems and the TSPLib benchmark datasets. Our findings illustrate that the MLHH framework offers a robust and innovative solution for tackling complex optimization challenges, paving the way for future research in this area.
For better reproducibility, we open source the code at \url{https://anonymous.4open.science/r/MHRE-BB53}.
Primary Area: optimization
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 7646
Loading