| INTERDISCIPLINARY PHYSICS AND RELATED AREAS OF SCIENCE AND TECHNOLOGY |
Prev
|
|
|
Evolutionary hypergraph dismantling via deep reinforcement learning |
| Junjie Qian(钱俊杰)†, Wenlan Wang(王文蓝)†, Hanyun Wang(王涵韵)†, Qiqi Wang(王萁淇)‡, Yao Zhang(张瑶)§, and Huijia Li(李慧嘉)¶ |
| School of Statistics and Data Science, AAIS, LPMC and KLMDASR, Nankai University, Tianjin 300074, China |
|
|
|
|
Abstract Assessing the vulnerability of complex systems requires effective hypergraph dismantling strategies, yet existing methods struggle with the dynamic nature of cascading failures and the rugged optimization landscapes of high-order networks. In this paper, we propose a novel framework: hypergraph dismantling via evolutionary deep reinforcement learning (HD-EDR). First, we model a realistic dismantling environment incorporating hyperdegree-based and residual-capacity-based load redistribution mechanisms. Second, we introduce a hybrid learning architecture that synergizes the global exploration of evolutionary strategies with the gradient-based exploitation of deep reinforcement learning. A bidirectional parameter synchronization mechanism is designed to prevent the agent from being trapped in local optima. Furthermore, we integrate an inductive encoder to capture the evolving high-order dependencies of the residual network in real time. Extensive experiments across nine real-world datasets demonstrate that our framework significantly outperforms state-of-the-art baselines, providing a highly effective and robust strategy for maximizing structural damage in high-order networks.
|
Received: 10 February 2026
Revised: 25 March 2026
Accepted manuscript online: 01 April 2026
|
|
PACS:
|
89.75.Hc
|
(Networks and genealogical trees)
|
| |
89.20.Ff
|
(Computer science and technology)
|
| |
07.05.Mh
|
(Neural networks, fuzzy logic, artificial intelligence)
|
| |
02.60.Pn
|
(Numerical optimization)
|
|
| Fund: Project supported by the National Natural Science Foundation of China (Grant Nos. 72571150 and 62306156). |
Corresponding Authors:
Qiqi Wang, Yao Zhang, Huijia Li
E-mail: qiqiwang@nankai.edu.cn;yaozhang@nankai.edu.cn;hjli@nankai.edu.cn
|
Cite this article:
Junjie Qian(钱俊杰), Wenlan Wang(王文蓝), Hanyun Wang(王涵韵), Qiqi Wang(王萁淇), Yao Zhang(张瑶), and Huijia Li(李慧嘉) Evolutionary hypergraph dismantling via deep reinforcement learning 2026 Chin. Phys. B 35 068901
|
[1] Calderoni F, Catanese S, De Meo P, Ficara A and Fiumara G 2020 Expert Syst. Appl. 161 113666
[2] Ma X Y, Zhou H J and Li Z Y 2021 Renew. Sustain. Energy Rev. 152 111646
[3] Zhang M Y, Huang T, Guo Z X and He Z G 2022 Physica A 607 128063
[4] Ullah A, Shoa J M, Yang Q L, Khan N, Bernard C M and Kumar R 2023 Expert Syst. Appl. 228 120326
[5] Albert R, Jeong H and Barabasi A L 2000 Nature 406 378
[6] Freeman L C 1979 Soc. Netw. 1 215
[7] Peng P, Fan T and Lu L 2024 Entropy 26 248
[8] Estrada E and Rodríguez-Velazquez J A 2005 Phys. Rev. E 71 056103
[9] Kapoor K, Sharma D and Srivastava J 2013 Proceedings of the 2013 IEEE 2nd Network Science Workshop (NSW) (West Point, NY, USA)
[10] Xie X W, Zhan X X, Zhang Z K and Liu C 2023 Chaos 33 013104
[11] Battiston F, Cencetti G, Iacopini I, Latora V, Lucas M, Patania A, Young J G and Petri G 2020 Phys. Rep. 874 1
[12] Qu H, Wang X, Song Y R, Ni W, Jiang G P and Sheng Q Z 2025 Proceedings of the 34th ACM International Conference on Information and Knowledge Management (Orlando, Florida, USA) p. 2450
[13] Jiang T, Liu K, Xiang B B, Zhang H F and Wang H 2025 Physica A 679 130988
[14] Motter A E and Lai Y C 2002 Phys. Rev. E 66 065102
[15] de Arruda G F, Petri G and Moreno Y 2020 Phys. Rev. Res. 2 023032
[16] Buldyrev S V, Parshani R, Paul G, Stanley H E and Havlin S 2010 Nature 464 1025
[17] Khalil E B, Dai H, Zhang Y, Dilkina B and Song L 2017 Proceedings of the Advances in Neural Information Processing Systems (NeurIPS) (Long Beach, CA, USA) p. 6348
[18] Manchanda S, Mittal A, Dhawan A, Medya S, Ranu S and Singh A 2020 Proceedings of the 34th International Conference on Neural Information Processing Systems (Virtual Conference) p. 20000
[19] Song N, Sheng W, Sun Y, Lin T, Wang Z, Xu Z, Yang F, Zhang Y and Li D 2025 Neurocomputing 618 129117
[20] Newman M E J 2003 SIAM Rev. 45 167
[21] Xu L, Ma L J, Lin Q Z, Li L J, Gong M G and Li J Q 2025 Inf. Sci. 698 121764
[22] Feng Y F, You H X, Zhang Z Z, Ji R R and Gao Y 2019 Proceedings of the AAAI Conference on Artificial Intelligence (Honolulu, Hawaii, USA) p. 3558
[23] Yadati N, Nimishakavi M, Yadav P, Nitin V, Louis A and Talukdar P 2019 Proceedings of the 33rd International Conference on Neural Information Processing Systems (Vancouver, Canada) p. 1511
[24] Hamilton W L, Ying R and Leskovec J 2017 Proceedings of the 31st International Conference on Neural Information Processing Systems (Long Beach, California, USA) p. 1025
[25] Colas C, Sigaud O and Oudeyer P Y 2018 Proceedings of the 35th International Conference on Machine Learning (Stockholm, Sweden) p. 1039
[26] Jia C X and Liu R R 2026 Reliab. Eng. Syst. Saf. 267 111850
[27] Zhou B, Ma X J, Ma F X and Gao S J 2023 AIMS Math. 8 4814
[28] Amburg I, Veldt N and Benson A R 2020 arXiv:2006.05863[math.PR]
[29] Benson A R, Abebe R, Schaub M T and Kleinberg J 2018 Proc. Natl. Acad. Sci. USA 115 E11221
[30] Yin H, Benson A R, Leskovec J and Gleich D F 2017 Proceedings of the 23rd ACM SIGKDD Int. Conf. Knowl. Discov. Data Min. (KDD ’17) (Halifax, NS, Canada) p. 555
[31] Petri G and Barrat A 2018 Phys. Rev. Lett. 121 228301
[32] Ren X L, Gleinig N, Helbing D and Antulov-Fantulin N 2019 Proc. Natl. Acad. Sci. USA 116 6554
[33] Morone F and Makse H A 2015 Nature 524 65
[34] Yan D, Xie W, Zhang Y, He Q and Yang Y 2022 IEEE Trans. Netw. Sci. Eng. 9 3302
[35] Fan C J, Zeng L, Sun Y Z and Liu Y Y 2020 Nat. Mach. Intell. 2 317 |
| No Suggested Reading articles found! |
|
|
Viewed |
|
|
|
Full text
|
|
|
|
|
Abstract
|
|
|
|
|
Cited |
|
|
|
|
Altmetric
|
|
blogs
Facebook pages
Wikipedia page
Google+ users
|
Online attention
Altmetric calculates a score based on the online attention an article receives. Each coloured thread in the circle represents a different type of online attention. The number in the centre is the Altmetric score. Social media and mainstream news media are the main sources that calculate the score. Reference managers such as Mendeley are also tracked but do not contribute to the score. Older articles often score higher because they have had more time to get noticed. To account for this, Altmetric has included the context data for other articles of a similar age.
View more on Altmetrics
|
|
|
