Solving Multiobjective Combinatorial Optimization via Learn to Improve Method
Primary Area: reinforcement learning
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.
Keywords: multi-objective combinatorial optimization, neural heuristic, learning to optimize, deep reinforcement learning
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Abstract: Recently, deep reinforcement learning (DRL) has been prevailing for solving multiobjective combinatorial optimization problems (MOCOPs). Most DRL methods are based on the "Learn to Construct" paradigm, where the trained model(s) can directly generate a set of approximate Pareto optimal solutions. However, these methods still suffer from insufficient proximity and poor diversity towards the true Pareto front. In this paper, we propose "Learn to Improve" (L2I), a learning-based improvement method for solving MOCOPs. We embed a weight-related policy network into multiobjective evolutionary algorithm (MOEA) frameworks to effectively guide the search direction. A shared baseline for proximal policy optimization is presented to reduce variance in model training. A quality enhancement mechanism is designed to further improve the Pareto set in model inference. Computational experiments conducted on two classic MOCOPs, i.e., multiobjective traveling salesman problem and multiobjective vehicle routing problem, indicate that our method achieves state-of-the-art results. Notably, our L2I module can be easily integrated into various MOEA frameworks such as NSGA-II, MOEA/D and MOGLS.
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: 5602
Loading