PILS: Exploring high-order neighborhoods by pattern mining and injection
作者：
Highlights：
• We introduce PILS: a new local search paradigm based on pattern mining and injection.
• Our method iteratively inserts a frequent pattern into the solution and completes it.
• This permits to detect high-order moves that would otherwise not be found.
• We apply it to the vehicle routing problem and propose an efficient completion algorithm.
• Our experiments demonstrate the large contribution of PILS to the search performance.
摘要
•We introduce PILS: a new local search paradigm based on pattern mining and injection.•Our method iteratively inserts a frequent pattern into the solution and completes it.•This permits to detect high-order moves that would otherwise not be found.•We apply it to the vehicle routing problem and propose an efficient completion algorithm.•Our experiments demonstrate the large contribution of PILS to the search performance.
论文关键词：Local search,Pattern mining,Combinatorial optimization,Vehicle routing problem
论文评审过程：Received 24 December 2019, Revised 10 December 2020, Accepted 18 March 2021, Available online 25 March 2021, Version of Record 3 May 2021.
论文官网地址：https://doi.org/10.1016/j.patcog.2021.107957
