当研究室で取り組んでいる研究に関する論文・講演などの情報（一部抜粋）です。
Here is a selection of our published papers on the research we’ve been working on in our laboratory.
Our Selected Papers and Keywords
2026
Yuki Aoyama, Kengo Katayama
Performance Comparison of Local Searches based on N-neighborhoods for the Job Shop Scheduling Problem Proceedings Article
In: GECCO '26 Companion: Proceedings of the Genetic and Evolutionary Computation Conference Companion, pp. 1277 - 1284, Association for Computing Machinery, 2026, ISBN: 979-8-4007-2488-6.
Abstract | BibTeX | タグ: Combinatorial optimization, Job shop scheduling problem, Local search, Neighborhood | Links:
@inproceedings{AOYAMA-GECCO-2026,
title = {Performance Comparison of Local Searches based on N-neighborhoods for the Job Shop Scheduling Problem},
author = {Yuki Aoyama and Kengo Katayama},
url = {https://dl.acm.org/doi/10.1145/3795101.3814697},
doi = {https://doi.org/10.1145/3795101.381469},
isbn = {979-8-4007-2488-6},
year = {2026},
date = {2026-08-13},
urldate = {2026-08-13},
booktitle = {GECCO '26 Companion: Proceedings of the Genetic and Evolutionary Computation Conference Companion},
pages = {1277 - 1284},
publisher = {Association for Computing Machinery},
abstract = {The Job-shop Scheduling Problem (JSSP) is a classical combinatorial optimization problem with significant implications for modern high-mix, low-volume manufacturing. Given its NP-hard complexity, local search based on critical-path-oriented neighborhood structures—specifically N5 through N8—has become a cornerstone of state-of-the-art metaheuristics. However, while existing literature focuses on their integration into complex frameworks like Tabu Search, their performance as standalone, basic local search (LS) is not yet fully understood. This obscures whether their efficacy is intrinsic to the neighborhood structures or dependent on the metaheuristic framework. Moreover, as these neighborhoods are nested, the incremental contribution of each additional move operator to objective function improvement has not been rigorously quantified. This paper presents a systematic empirical evaluation of the neighborhoods to elucidate their standalone effectiveness. Through extensive computational experiments on 28 standard benchmark instances, we comprehensively evaluate the impact of typical improvement strategies and neighborhood evaluation orders on search performance. Our findings demonstrate that the N8-based LS consistently outperforms other structures in terms of makespan reduction. Furthermore, a contribution analysis substantiates the practical utility of the nested neighborhood design, establishing the N8 structure as a potent intensification tool for manufacturing scheduling.},
keywords = {Combinatorial optimization, Job shop scheduling problem, Local search, Neighborhood},
pubstate = {published},
tppubtype = {inproceedings}
}
The Job-shop Scheduling Problem (JSSP) is a classical combinatorial optimization problem with significant implications for modern high-mix, low-volume manufacturing. Given its NP-hard complexity, local search based on critical-path-oriented neighborhood structures—specifically N5 through N8—has become a cornerstone of state-of-the-art metaheuristics. However, while existing literature focuses on their integration into complex frameworks like Tabu Search, their performance as standalone, basic local search (LS) is not yet fully understood. This obscures whether their efficacy is intrinsic to the neighborhood structures or dependent on the metaheuristic framework. Moreover, as these neighborhoods are nested, the incremental contribution of each additional move operator to objective function improvement has not been rigorously quantified. This paper presents a systematic empirical evaluation of the neighborhoods to elucidate their standalone effectiveness. Through extensive computational experiments on 28 standard benchmark instances, we comprehensively evaluate the impact of typical improvement strategies and neighborhood evaluation orders on search performance. Our findings demonstrate that the N8-based LS consistently outperforms other structures in terms of makespan reduction. Furthermore, a contribution analysis substantiates the practical utility of the nested neighborhood design, establishing the N8 structure as a potent intensification tool for manufacturing scheduling.
2005
Kengo Katayama, Akihiro Hamamoto, Hiroyuki Narihisa
An effective local search for the maximum clique problem Journal Article
In: Information Processing Letters, vol. 95, no. 5, pp. 503-511, 2005, ISSN: 0020-0190.
Abstract | BibTeX | タグ: Combinatorial optimization, Graph algorithms, Maximum clique problem, Neighborhood, Variable depth search | Links:
@article{KATAYAMA-IPL-2005,
title = {An effective local search for the maximum clique problem},
author = {Kengo Katayama and Akihiro Hamamoto and Hiroyuki Narihisa},
url = {https://www.sciencedirect.com/science/article/pii/S0020019005001444},
doi = {https://doi.org/10.1016/j.ipl.2005.05.010},
issn = {0020-0190},
year = {2005},
date = {2005-01-01},
urldate = {2005-01-01},
journal = {Information Processing Letters},
volume = {95},
number = {5},
pages = {503-511},
abstract = {We propose a variable depth search based algorithm, called k-opt local search (KLS), for the maximum clique problem. KLS efficiently explores the k-opt neighborhood defined as the set of neighbors that can be obtained by a sequence of several add and drop moves that are adaptively changed in the feasible search space. Computational results on DIMACS benchmark graphs indicate that KLS is capable of finding considerably satisfactory cliques with reasonable running times in comparison with those of state-of-the-art metaheuristics.},
keywords = {Combinatorial optimization, Graph algorithms, Maximum clique problem, Neighborhood, Variable depth search},
pubstate = {published},
tppubtype = {article}
}
We propose a variable depth search based algorithm, called k-opt local search (KLS), for the maximum clique problem. KLS efficiently explores the k-opt neighborhood defined as the set of neighbors that can be obtained by a sequence of several add and drop moves that are adaptively changed in the feasible search space. Computational results on DIMACS benchmark graphs indicate that KLS is capable of finding considerably satisfactory cliques with reasonable running times in comparison with those of state-of-the-art metaheuristics.