Neural Deconstruction Search for Vehicle Routing Problems
Andr\'e Hottung, Paula Wong-Chung, Kevin Tierney
TL;DR
This paper introduces a neural deconstruction search method for vehicle routing problems that iteratively deconstructs and reconstructs solutions, outperforming traditional sequential construction approaches and matching state-of-the-art methods.
Contribution
It proposes a novel iterative search framework using neural policies for deconstruction and reconstruction, challenging the conventional sequential solution construction paradigm.
Findings
Matches or surpasses state-of-the-art methods across multiple vehicle routing problems
Effective collaboration between neural policy and greedy insertion algorithm
Applicable to various problem sizes with high solution quality
Abstract
Autoregressive construction approaches generate solutions to vehicle routing problems in a step-by-step fashion, leading to high-quality solutions that are nearing the performance achieved by handcrafted operations research techniques. In this work, we challenge the conventional paradigm of sequential solution construction and introduce an iterative search framework where solutions are instead deconstructed by a neural policy. Throughout the search, the neural policy collaborates with a simple greedy insertion algorithm to rebuild the deconstructed solutions. Our approach matches or surpasses the performance of state-of-the-art operations research methods across three challenging vehicle routing problems of various problem sizes.
Peer Reviews
Decision·Submitted to ICLR 2025
- Originality: The paper proposes a unique deconstruction-reconstruction framework for VRPs, leveraging a reinforcement learning policy to optimize solution quality in a novel way. - Quality: The experiments are thorough, showing significant performance improvements across multiple VRP benchmarks and validating the approach’s robustness. - Clarity: The paper is well-organized, with clear explanations and visuals that make the deconstruction and reconstruction process easy to understand.
- The paper introduces a deconstruction and re-insertion heuristic for VRP improvement, but it lacks a clear comparison with well-known heuristics, such as the 2-opt method, which also iteratively refines solutions. Providing an explanation of how the proposed approach differs from 2-opt would clarify the advantages of using a learning-based method for the deconstruction-recreation process. - While the paper presents NDS as a novel learning-based improvement heuristic (Costa, 2020), it does not
1) The paper is technically well-written and easy to follow 2) The presented experiments are of sufficient breadth 3) The proposed methodology opens an interesting avenue for algorithm design that deviates from existing approaches that usually focus on using neural policies to construct solutions in a step-by-step fashion. 4) The results show promising performance compared to existing methods.
As indicated in my summary, the authors are currently (significantly) overselling the contribution of the proposed method. This relates to the fact that numerical experiments are - to some extent - comparing apples with oranges due to limiting the computation times of the studied algorithms instead of using a proper performance-based stopping criterion. This experimental design choice clearly favors the search technique proposed by the authors, as the neural deconstruction works instantaneously
1. The paper writing is fluent. 2. Experiments show outstanding performances on three VRPs.
1. From my point of view, the framework of this paper is mainly based on SISRs. This paper replaces the heuristic destruction process of SISRs with a learning-based destruction strategy. The greedy insertion with simulated annealing in this paper is essentially the same as the heuristic reconstruction process "greedy insertion with blinks" in SISRs, both of which have the probability of accepting differential solutions to jump out of the local optimum. Based on the above observations, I think th
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsManufacturing Process and Optimization