M<sup>3</sup>Net: Meta-Reinforcement Learning-Based Open-Set Domain Generalization of Hyperspectral Image Classification Model

Yuhu Cheng; Wei Zhang; C. L. Philip Chen; Xuesong Wang

doi:10.1109/JAS.2025.125981

IEEE/CAA Journal of Automatica Sinica

JCR Impact Factor: 19.2, Top 1 (SCI Q1)

CiteScore: 28.2, Top 1% (Q1)
Google Scholar h5-index: 95， TOP 5

Turn off MathJax

Article Contents

Article Navigation > IEEE/CAA Journal of Automatica Sinica > 2026 > In Press, Accepted Manuscript

Y. Cheng, W. Zhang, C. L. Philip Chen, and X. Wang, “M3Net: Meta-reinforcement learning-based open-set domain generalization of hyperspectral image classification model,” IEEE/CAA J. Autom. Sinica, early access, 2026. doi: 10.1109/JAS.2025.125981

Citation:

Y. Cheng, W. Zhang, C. L. Philip Chen, and X. Wang, “M³Net: Meta-reinforcement learning-based open-set domain generalization of hyperspectral image classification model,” IEEE/CAA J. Autom. Sinica, early access, 2026. doi: 10.1109/JAS.2025.125981

Citation:

PDF( 2861 KB)

M³Net: Meta-Reinforcement Learning-Based Open-Set Domain Generalization of Hyperspectral Image Classification Model

doi: 10.1109/JAS.2025.125981

Funds: This work was funded by the National Natural Science Foundation of China (62373364, 62573416), and by the Key Research and Development Program of Jiangsu Province under

More Information

Author Bio:
Yuhu Cheng (Senior Member, IEEE) received the Ph.D. degree in Control Science and Engineering from Institute of Automation, Chinese Academy of Sciences in 2005. Prof. Cheng is currently a Professor with the School of Information and Control Engineering, China University of Mining and Technology. His main research interests include machine learning and intelligent systems

Wei Zhang (Student Member, IEEE) received the M.S. degree in Control Science and Engineering from China University of Mining and Technology in 2024. He is currently pursuing the Ph.D. degree in Control Science and Engineering with the School of Information and Control Engineering, China University of Mining and Technology. His current research interests include machine learning and pattern recognition

C. L. Philip Chen (Life Fellow, IEEE) is the Chair Professor and Dean with the College of Computer Science and Engineering, South China University of Technology. Being a Program Evaluator of the Accreditation Board of Engineering and Technology Education (ABET) in the U.S., for computer engineering, electrical engineering, and software engineering programs, he successfully architects the University of Macau’s Engineering and Computer Science programs receiving accreditations from Washington/Seoul Accord through Hong Kong Institute of Engineers (HKIE), of which is considered as his utmost contribution in engineering/computer science education for Macau as the former Dean of the Faculty of Science and Technology. He is a Fellow of IEEE, AAAS, IAPR, CAA, and HKIE; a member of Academia Europaea (AE), European Academy of Sciences and Arts (EASA), and International Academy of Systems and Cybernetics Science (IASCYS). He received IEEE Norbert Wiener Award in 2018 for his contribution to systems and cybernetics, and machine learning. His current research interests include cybernetics, systems, and computational intelligence. Prof. Chen was a recipient of the 2016 Outstanding Electrical and Computer Engineers Award from his alma mater, Purdue University (in 1988), after he graduated from the University of Michigan at Ann Arbor, Ann Arbor, MI, USA in 1985. He was the IEEE Systems, Man, and Cybernetics Society President from 2012 to 2013, the Editor-in-Chief of IEEE Transactions on Systems, Man, and Cybernetics: Systems (2014−2019), and the Editor-in-Chief of IEEE Transactions on Cybernetics (2019−2020), and an Associate Editor of IEEE Transactions on Fuzzy Systems. He was the Chair of TC 9.1 Economic and Business Systems of International Federation of Automatic Control from 2015 to 2017

Xuesong Wang (Member, IEEE) received the Ph.D. degree in Control Science and Engineering from China University of Mining and Technology in 2002. She is currently the Dean of Engineering Research Center of Intelligent Control for Underground Space, Ministry of Education; and a Professor with the School of Information and Control Engineering, China University of Mining and Technology. Her research interests include machine learning and artificial intelligence. Prof. Wang is currently a Senior Editor of IEEE Transactions on Systems, Man, and Cybernetics: Systems, an Associate Editor of IEEE Transactions on Neural Networks and Learning Systems, IEEE Transactions on Systems, Man, and Cybernetics: Systems, IEEE Signal Processing Letters, International Journal of Machine Learning and Cybernetics, Acta Automatica Sinica, and Acta Electronica Sinica
Corresponding author: Xuesong Wang, e-mail: wangxuesongcumt@163.com
Received Date: 2025-05-16
Revised Date: 2025-09-09
Accepted Date: 2025-10-11

Available Online: 2026-03-05

Abstract

Abstract

Hyperspectral image (HSI) classification models face dual challenges in open-set domain generalization: limited generalization ability due to unseen-domain shifts, and the need for unknown class recognition that breaks the closed-set assumption of traditional models. To address these challenges, we propose the Markov meta-Mamba network (M³Net), which provides a meta-reinforcement learning-based solution for open-set domain generalization of HSI classification model. Specifically, a meta-task construction mechanism is proposed, treating source-domain background pixels as virtual unknown classes to simulate open-set HSI classification tasks during training, thereby providing task support for meta-reinforcement learning. Then, the open-set HSI classification task is reconstructed as a Markov decision process. By leveraging reinforcement learning’s multi-step temporal credit assignment, non-causal factor sensitivity is suppressed, improving the model’s cross-domain generalization performance. Finally, the theoretical linkage between Mamba and meta-learning is established, demonstrating that Mamba inherently operates as a meta-learner when processing task sequences. Building on this, a Mamba-based meta-task embedding framework is designed, where shared meta-parameters and task-specific parameters are jointly optimized to achieve cross-task knowledge induction across open-set HSI classification tasks, thereby enhancing the model’s generalization capability for unseen open-set tasks. Experiments on three cross-domain hyperspectral image datasets show that M³Net has achieved the most competitive performance in the open-set domain generalization.
- Hyperspectral image classification,
- Markov decision process,
- meta-learning,
- open-set domain generalization,
- reinforcement learning

FullText(HTML)

References(42)

References

[1]	K. Gao, B. Liu, X. Yu, and A. Z. Yu, “Unsupervised meta learning with multiview constraints for hyperspectral image small sample set classification,” IEEE Trans. Image Process., vol. 31, pp. 3449–3462, May 2022. doi: 10.1109/TIP.2022.3169689
[2]	M. Zhang, X. Zhao, W. Li, Y. Zhang, R. Tao, and Q. Du, “Cross-scene joint classification of multisource data with multilevel domain adaption network,” IEEE Trans. Neural Netw. Learn. Syst., vol. 35, no. 8, pp. 11514–11526, Aug. 2024. doi: 10.1109/TNNLS.2023.3262599
[3]	J. Li, Z. Zhang, R. Song, Y. Li, and Q. Du, “SCFormer: Spectral coordinate transformer for cross-domain few-shot hyperspectral image classification,” IEEE Trans. Image Process., vol. 33, pp. 840–855, Jan. 2024. doi: 10.1109/TIP.2024.3351443
[4]	D. Wang, J. Zhang, B. Du, L. Zhang, and D. Tao, “DCN-T: Dual context network with transformer for hyperspectral image classification,” IEEE Trans. Image Process., vol. 32, pp. 2536–2551, Apr. 2023. doi: 10.1109/TIP.2023.3270104
[5]	J. Yang, B. Du, and L. Zhang, “From center to surrounding: An interactive learning framework for hyperspectral image classification,” ISPRS J Photogramm. Remote Sens., vol. 197, pp. 145–166, Mar. 2023. doi: 10.1016/j.isprsjprs.2023.01.024
[6]	D. Wang, M. Hu, Y. Jin, Y. Miao, J. Yang, and Y. Xu, “HyperSIGMA: Hyperspectral intelligence comprehension foundation model,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 47, no. 8, pp. 6427–6444, Aug. 2025. doi: 10.1109/TPAMI.2025.3557581
[7]	Y. Zhang, W. Li, W. Sun, R. Tao, and Q. Du., “Single-source domain expansion network for cross-scene hyperspectral image classification,” IEEE Trans. Image Process., vol. 32, pp. 1498–1512, Feb. 2023. doi: 10.1109/TIP.2023.3243853
[8]	B. Qin, S. Feng, C. Zhao, B. Xi, W. Li, and R. Tao, “FDGNet: Frequency disentanglement and data geometry for domain generalization in cross-scene hyperspectral image classification,” IEEE Trans. Neural Netw. Learn. Syst., vol. 36, no. 6, pp. 10297–10310, Jun. 2025. doi: 10.1109/TNNLS.2024.3445136
[9]	N. Li, X. Song, Y. Liu, W. Zhu, C. Li, W. Zhang, and Y. Quan, “Semi-supervised graph constraint dual classifier network with unknown class feature learning for hyperspectral image open-set classification,” IEEE Geosci. Remote Sens. Lett., vol. 22, pp. 5504005, Apr. 2025.
[10]	Y. Zhang, M. Zhang, W. Li, S. Wang, and R. Tao, “Language-aware domain generalization network for cross-scene hyperspectral image classification,” IEEE Trans. Geosci. Remote Sens., vol. 61, p. 5501312, Jan. 2023. doi: 10.2139/ssrn.4937361
[11]	B. Qin, S. Feng, C. Zhao, W. Li, R. Tao, and J. Zhou, “Language-enhanced dual-level contrastive learning network for open-set hyperspectral image classification,” IEEE Trans. Geosci. Remote Sens., vol. 63, pp. 5508114, Mar. 2025.
[12]	Y. Du, X. Li, L. Shi, F. Li, and T. Xu, “A prototype network for hyperspectral image open-set classification based on feature invariance and weighted Pearson distance measurement,” IEEE Trans. Geosci. Remote Sens., vol. 62, p. 5506917, Jan. 2024. doi: 10.1109/tgrs.2024.3359311
[13]	H. Wang, X. Liu, Z. Qiao, and H. Tao, “Inducing causal meta-knowledge from virtual domain: Causal meta-generalization for hyperspectral domain generalization,” IEEE Trans. Geosci. Remote Sens., vol. 62, pp. 5538416, Nov. 2024.
[14]	X. Zhou, X. Zheng, T. Shu, W. Liang, K. I. K. Wang, L. Qi, S. Shimizu, and Q. Jin, “Information theoretic learning-enhanced dual-generative adversarial networks with causal representation for robust OOD generalization,” IEEE Trans, Neural Netw. Learn. Syst., vol. 36, no. 2, pp. 2066–2079, Feb. 2025. doi: 10.1109/TNNLS.2023.3330864
[15]	F. Lv, J. Liang, S. Li, B. Zang, C. H. Liu, Z. Wang, and D. Liu, “Causality inspired representation learning for domain generalization,” in Proc IEEE/CVF Conf. Computer Vision and Pattern Recognition, New Orleans, LA, USA, 2022, pp. 8046−8056.
[16]	J. Li, Z. Zhang, R. Song, H. Xu, Y. Li, and Q. Du, “Contrastive MLP network based on adjacent coordinates for cross-domain zero-shot hyperspectral image classification,” IEEE Trans. Circuits Syst. Video Technol., vol. 35, no. 8, pp. 8377–8390, Aug. 2025. doi: 10.1109/TCSVT.2025.3549365
[17]	X. Wang, S. Wang, X. Liang, D. Zhao, J. Huang, X. Xu, B. Dai, and Q. Miao, “Deep reinforcement learning: A survey,” IEEE Trans. Neural Netw. Learn. Syst., vol. 35, no. 4, pp. 5064–5078, Apr. 2024. doi: 10.1109/TNNLS.2022.3207346
[18]	S. Khairy, P. Balaprakash, and L. X. Cai, “A gradient-aware search algorithm for constrained Markov decision processes,” IEEE Trans. Neural Netw. Learn. Syst., vol. 35, no. 12, pp. 18914–18921, Dec. 2024. doi: 10.1109/TNNLS.2023.3315598
[19]	T. Zhang, Z. Lin, Y. Wang, D. Ye, Q. Fu, W. Yang, X. Wang, B. Liang, B. Yuan, and X. Li, “Dynamics-adaptive continual reinforcement learning via progressive contextualization,” IEEE Trans. Neural Netw. Learn. Syst., vol. 35, no. 10, pp. 14588–14602, Oct. 2024. doi: 10.1109/TNNLS.2023.3280085
[20]	G. Chen, P. Peng, X. Wang, and Y. Tian, “Adversarial reciprocal points learning for open set recognition,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 44, no. 11, pp. 8065–8081, Nov. 2022.
[21]	B. Wang, Y. Xu, Z. Wu, S. Zheng, Z. Wei, and J. Chanussot, “Hyperspectral images single-source domain generalization based on nonlinear sample generation,” IEEE Trans. Geosci. Remote Sens., vol. 62, pp. 5516613, Apr. 2024.
[22]	Y. Sun, B. Liu, R. Wang, P. Zhang, and M. Dai, “Spectral-spatial MLP-like network with reciprocal points learning for open-set hyperspectral image classification,” IEEE Trans. Geosci. Remote Sens., vol. 61, pp. 5513218, May 2023.
[23]	T. Hospedales, A. Antoniou, P. Micaelli, and A. Storkey, “Meta-learning in neural networks: A survey,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 44, no. 9, pp. 5149–5169, Sep. 2022.
[24]	A. Gu, and T. Dao, “Mamba: Linear-time sequence modeling with selective state spaces,” arXiv preprint arXiv: 2312.00752, 2024.
[25]	B. Zhang, Y. Chen, S. Xiong, and X. Lu, “Hyperspectral image classification via cascaded spatial cross-attention network,” IEEE Trans. Image Process., vol. 34, pp. 899–913, Jan. 2025. doi: 10.1109/TIP.2025.3533205
[26]	M. Ahmad, M. Mazzara, S. Distefano, A. M. Khan, M. H. F. Butt, and D. Hong, “PolicyMamba: Localized policy attention with state space model for land cover classification,” IEEE Trans. Neural Netw. Learn. Syst., vol. 36, no. 10, pp. 17814–17825, Oct. 2025. doi: 10.1109/TNNLS.2025.3586836
[27]	Y. Shao, W. Wu, X. You, C. Gao, and N. Sang, “Improving the generalization of MAML in few-shot classification via bi-level constraint,” IEEE Trans. Circuits Syst. Video Technol., vol. 33, no. 7, pp. 3284–3295, Jul. 2023. doi: 10.1109/TCSVT.2022.3232717
[28]	S. Baik, M. Choi, J. Choi, H. Kim, and K. M. Lee, “Learning to learn task-adaptive hyperparameters for few-shot learning,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 46, no. 3, pp. 1441–1454, Mar. 2024. doi: 10.1109/TPAMI.2023.3261387
[29]	P. Tang, X. Luo, and J. Woodcock, “Auto-encoding neural tucker factorization,” IEEE Trans. Knowl. Data Eng., vol. 37, no. 10, pp. 5795–5807, Oct. 2025. doi: 10.1109/TKDE.2025.3590198
[30]	X. Luo, H. Wu, and Z. Li, “Neulft: A novel approach to nonlinear canonical polyadic decomposition on high-dimensional incomplete tensors,” IEEE Trans. Knowl. Data Eng., vol. 35, no. 6, pp. 6148–6166, Jun. 2023. doi: 10.1109/tkde.2022.3176466
[31]	J. Bai, W. Shi, Z. Xiao, T. A. A. Ali, F. Ye, and L. Jiao, “Achieving better category separability for hyperspectral image classification: A spatial-spectral approach,” IEEE Trans. Neural Netw. Learn. Syst., vol. 35, no. 7, pp. 9621–9635, Jul. 2024. doi: 10.1109/TNNLS.2023.3235711
[32]	J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, “Proximal policy optimization algorithms,” arXiv preprint arXiv: 1707.06347, 2017.
[33]	M. S. Ausin, H. Azizsoltani, S. Ju, Y. J. Kim, and M. Chi, “InferNet for delayed reinforcement tasks: Addressing the temporal credit assignment problem,” in Proc. IEEE Int. Conf. Big Data, Orlando, FL, USA, 2021, pp. 1337−1348.
[34]	Y. Wu, S. Liao, X. Liu, Z. Li, and R. Lu, “Deep reinforcement learning on autonomous driving policy with auxiliary critic network,” IEEE Trans. Neural Netw. Learn. Syst., vol. 34, no. 7, pp. 3680–3690, Jul. 2023. doi: 10.1109/TNNLS.2021.3116063
[35]	M. Zhang, S. Zhang, X. Wu, Z. Shi, X. Deng, E. Wu, and X. Q. Xu, “Efficient reinforcement learning with the novel N-step method and V-network,” IEEE Trans. Cybern., vol. 54, no. 10, pp. 6048–6057, Oct. 2024. doi: 10.1109/TCYB.2024.3401014
[36]	A. Bendale and T. E. Boult, “Towards open set deep networks,” in Proc. IEEE Conf. Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 2016, pp. 1563−1572.
[37]	Y. Li, Y. Luo, L. Zhang, Z. Wang, and B. Du, “MambaHSI: Spatial-spectral Mamba for hyperspectral image classification,” IEEE Trans. Geosci. Remote Sens., vol. 62, pp. 5524216, Jul. 2024.
[38]	X. Yang, W. Cao, Y. Lu, and Y. Zhou, “Hyperspectral image transformer classification networks,” IEEE Trans. Geosci. Remote Sens., vol. 60, pp. 5528715, May 2022.
[39]	P. R. M. Júnior, R. M. de Souza, R. de O. Werneck, B. V. Stein, D. V. Pazinato, W. R. de Almeida, O. A. B. Penatti, R. D. S. Torres, and A. Rocha, “Nearest neighbors distance ratio open-set classifier,” Mach. Learn., vol. 106, no. 3, pp. 359–386, Mar. 2017. doi: 10.1007/s10994-016-5610-8
[40]	Z. Xie, P. Duan, W. Liu, X. Kang, X. Wei, and S. Li, “Feature consistency-based prototype network for open-set hyperspectral image classification,” IEEE Trans. Neural Netw. Learn. Syst., vol. 35, no. 7, pp. 9286–9296, Jul. 2024. doi: 10.1109/TNNLS.2022.3232225
[41]	J. Jang and C. O. Kim, “Collective decision of one-vs-rest networks for open-set recognition,” IEEE Trans. Neural Netw. Learn. Syst., vol. 35, no. 2, pp. 2327–2338, Feb. 2024. doi: 10.1109/TNNLS.2022.3189996
[42]	J. Liu, J. Tian, W. Han, Z. Qin, Y. Fan, and J. Shao, “Learning multiple Gaussian prototypes for open-set recognition,” Inf. Sci., vol. 626, pp. 738–753, May 2023. doi: 10.1016/j.ins.2023.01.062

Supplements(0)

Cited By

Proportional views

Proportional views

通讯作者: 陈斌, bchen63@163.com

1.
沈阳化工大学材料科学与工程学院沈阳 110142

Figures(11) / Tables(4)

Get Citation

PDF

XML

Article Metrics

Article views (23) PDF downloads(3)

M3Net: Meta-Reinforcement Learning-Based Open-Set Domain Generalization of Hyperspectral Image Classification Model

doi: 10.1109/JAS.2025.125981

Abstract

References

Proportional views

Catalog

通讯作者: 陈斌, bchen63@163.com

Article Metrics

Export File

Citation

Format

Content

M³Net: Meta-Reinforcement Learning-Based Open-Set Domain Generalization of Hyperspectral Image Classification Model