HMCNet: یک چارچوب نوین برای خوشه‌بندی چندنما در شرایط ناقص بودن نماها
سندس بهادری
1
(
دانشکده فنی مهندسی، واحد ایلام، دانشگاه آزاد اسلامی، ایلام، ایران
)
کلید واژه: تابع زیان احتمالی, خوشه‌بندی چندنما, ماژول توجه, نماهای ناقص, نهفته¬سازی مبتنی بر کاهش ابعاد غیرخطی.,
چکیده مقاله :
خوشه‌بندی داده‌های چندنما در حضور نماهای ناقص یکی از چالش‌های مهم در یادگیری ماشین و داده‌کاوی است. روش‌های موجود غالباً در مواجهه با نرخ‌های بالای ناقص بودن نماها دچار افت عملکرد می‌شوند و قادر به استخراج ویژگی‌های مشترک و سازگار میان نماها نیستند. با توجه به آمار و نمونه‌های عینی بیان‌شده که نشان‌دهنده شیوع بالای ناقص‌بودن نماها در کاربردهای واقعی (با نرخ‌های بین 10% تا 50%) و افت شدید عملکرد روش‌های سنتی در این شرایط است، در این پژوهش مدل HMCNet معرفی شده است. این مدل با بهره‌گیری از ترکیب ماژول توجه، کاهش ابعاد غیرخطی UMAP و تابع زیان ترکیبی مبتنی بر احتمال، قادر است ساختارهای محلی داده‌ها را حفظ کرده و سازگاری میان نماها را حتی در حضور نرخ‌های بالای ناقص‌بودن (تا 50%) تقویت نماید. برای ارزیابی، آزمایش‌های گسترده‌ای بر روی دادگانها انجام شد. نتایج نشان داد که مدل پیشنهادی در معیارهای ACC، NMI و Purity به‌طور پایدار عملکردی برتر نسبت به روش‌های پیشرفته دارد. همچنین تحلیل‌های آزمایش‌های حذف اهمیت هر مؤلفه را در بهبود عملکرد مدل تأیید کردند. به‌طور کلی، یافته‌های این پژوهش نشان می‌دهند که HMCNet می‌تواند به‌عنوان یک چارچوب عمومی و قدرتمند برای خوشه‌بندی داده‌های چندنما در شرایط ناقص بودن نماها مورد استفاده قرار گیرد و مسیرهای جدیدی برای کاربردهای واقعی در حوزه‌های مختلف علمی و صنعتی فراهم سازد.
چکیده انگلیسی :
Clustering multi-view data in the presence of incomplete views is one of the major challenges in machine learning and data mining. Existing methods often suffer from performance degradation when faced with high rates of incomplete views and are unable to extract common and consistent features between views. Considering the statistics and objective examples presented, which indicate the high prevalence of incomplete views in real applications (with rates between 10% and 50%) and the severe performance degradation of traditional methods in these conditions, the HMCNet model is introduced in this study. By utilizing the combination of attention module, nonlinear UMAP dimensionality reduction, and probability-based composite loss function, this model is able to preserve the local structures of the data and enhance the consistency between views even in the presence of high incompleteness rates (up to 50%). Extensive experiments were conducted on datasets for evaluation. The results showed that the proposed model consistently outperforms state-of-the-art methods in ACC, NMI, and Purity metrics. Also, the analyses of deletion experiments confirmed the importance of each component in improving the model performance. Overall, the findings of this research show that HMCNet can be used as a general and powerful framework for clustering multi-view data under incomplete view conditions and provide new avenues for real-world applications in various scientific and industrial fields.
[1] K. Berahmand, S. Bahadori, M. N. Abadeh, Y. Li, and Y. Xu, "SDAC-DA: Semi-supervised deep attributed clustering using dual autoencoder," IEEE Trans. on Knowledge and Data Engineering, vol. 36, no. 11, pp. 6989-7002, 2024.
[2] X.-L. Li, M.-S. Chen, C.-D. Wang, and J.-H. Lai, "Refining graph structure for incomplete multi-view clustering IEEE Trans. on Neural Networks and Learning Systems, vol. 35, no. 2, pp. 2300-2313, 2022.
[3] Y. Chen, X. Zhao, Z. Zhang, Y. Liu, J. Su, and Y. Zhou, "Tensor learning meets dynamic anchor learning: From complete to incomplete multiview clustering," IEEE Trans. on Neural Networks and Learning Systems, vol. 35, no. 11, pp. 15332-15345, 2023.
[4] J. Wen, et al., "Structural deep incomplete multi-view clustering network," in Proc, of the 30th ACM Int. Conf. on Information & Knowledge Management, pp. 3538-3542, Virtual Event Queensland, Australia, 1-5 Nov. 2021.
[5] L. McInnes, J. Healy, and J. Melville, Umap: Uniform Manifold Aapproximation and Projection for Dimension Reduction, arXiv preprint arXiv:1802.03426, 2018.
[6] C. Leng, H. Zhang, G. Cai, I. Cheng, and A. Basu, "Graph regularized Lp smooth non-negative matrix factorization for data representation," IEEE/CAA Journal of Automatica Sinica, vol. 6, no. 2, pp. 584-595, 2019.
[7] G. A. Khan, "Complementary incomplete weighted concept factorization methods for multi-view clustering," Knowledge and Information Systems, vol. 66, no. 12, pp. 7469-7494, 2024.
[8] J. Wen, et al., "Graph regularized and feature aware matrix factorization for robust incomplete multi-view clustering," IEEE Trans. on Circuits and Systems for Video Technology, vol. 34, no. 5, pp. 3728-3741, 2023.
[9] Y. Ban, Y. Cai, and Z. Huang, "Autoencoder-like non-negative matrix factorization with dual-graph constraints for multi-view clustering," International Journal of Machine Learning and Cybernetics, pp. 1-16, 2025.
[10] Y. Khalafaoui, N. Grozavu, B. Matei, and L. -W. Goix, "Multi-modal multi-view clustering based on non-negative matrix factorization," in Proc. IEEE Symp. Series on Computational Intelligence, pp. 1386-1391, Singapore, Singapore, 4-7 Dec. 2022.
[11] Z. Shu, Q. Long, L. Zhang, Z. Yu, and X.-J. Wu, "Robust graph regularized NMF with dissimilarity and similarity constraints for ScRNA-seq data clustering," Journal of Chemical Information and Modeling, vol. 62, no. 23, pp. 6271-6286, 2022.
[12] W. Lv, C. Zhang, H. Li, X. Jia, and C. Chen, "Joint projection learning and tensor decomposition-based incomplete multiview clustering," IEEE Trans. on Neural Networks and Learning Systems, vol. 35, no. 12, pp. 17559-17570, Dec. 2024.
[13] W. Li and C. Zhu, "Cross-view attention with adversarial learning for incomplete multi-view clustering," Multimedia Systems, vol. 31, no. 5, p. 347, 2025.
[14] T. Wu, S. Feng, and J. Yuan, "Low-rank kernel tensor learning for incomplete multi-view clustering," in Proc. of the AAAI Conf. on Artificial Intelligence, pp. 15952-15960, 20-27 Feb. 2024.
[15] M. Briscik, G. Tazza, L. Vidács, M. -A. Dillies, and S. Déjean, "Supervised multiple kernel learning approaches for multi-omics data integration," BioData Mining, vol. 17, Article ID: 53, 2024.
[16] Z. Wen, et al., MultiKernelBench: A Multi-Platform Benchmark for Kernel Generation, arXiv eprints, pp. arXiv–2507, 2025.
[17] R. Zhang, Y. Li, Y. Gui, D. J. Armaghani, and M. Yari, "A stacked deep multi-kernel learning framework for blast induced flyrock prediction," International Journal of Rock Mechanics and Mining Sciences, vol. 177, Article ID: 105741, 2024.
[18] S. Wang, et al., "Highly-efficient incomplete large-scale multi-view clustering with consensus bipartite graph," in Proc. of the IEEE/CVF Conf. on Computer Vision and Pattern Recognition, pp. 9776-9785, New Orleans, LA, USA, 18-24 Jun. 2022.
[19] J. Wen, Z. Zhang, Y. Xu, and Z. Zhong, "Incomplete multi-view clustering via graph regularized matrix factorization," in Proc. of the European Conf. on Computer Vision Workshops, pp. 593-608, Munich, Germany, 8-14 Sept. 2018.
[20] J. Cao, J. Hu, and R. Zhang, "Adaptive graph learning algorithm for incomplete multi-view clustered image segmentation," Engineering Applications of Artificial Intelligence, vol. 139, Article ID: 109264, 2025.
[21] Z. Shu, B. Li, C. Hu, Z. Yu, and X.-J. Wu, "Robust dual-graph regularized deep matrix factorization for multi-view clustering," Neural Processing Letters, vol. 55, no. 5, pp. 6067-6087, 2023.
[22] D. Zhuang, et al., "Robust graph contrastive learning for incomplete multi-view clustering," in Proc. of the Thirty-Fourth Int. Joint Conf. on Artificial Intelligence, pp. 7282-7290, Montreal, Canada, 16-22 Aug. 2025.
[23] W. Xia, Q. Wang, Q. Gao, X. Zhang, and X. Gao, "Self-supervised graph convolutional network for multi-view clustering," IEEE Trans. on Multimedia, vol. 24, pp. 3182-3192, 2021.
[24] Y. Wen et al., "Scalable incomplete multi-view clustering with structure alignment," in Proc. of the 31st ACM Int. Conf. on Multimedia, pp. 3031-3040, Ottawa, Canada, 29 Oct.- 3 Nov 2023.
[25] J. Xu, et al., "Adaptive feature projection with distribution alignment for deep incomplete multi-view clustering," IEEE Trans. on Image Processing, vol. 32, pp. 1354-1366, 2023.
[26] W. Yan, K. Liu, W. Zhou, and C. Tang, "Deep incomplete multi-view clustering via dynamic imputation and triple alignment with dual optimization," IEEE Trans. on Circuits and Systems for Video Technology, vol. 35, no. 4, 3250 – 3261, Apr. 2024.
[27] Y. Li and B. Wu, "Joint generative and alignment adversarial learning for robust incomplete multi-view clustering," Neural Networks, vol. 194, Article ID: 108141, Feb. 2026.
[28] J. Yin, P. Wang, S. Sun, and Z. Zheng, "Incomplete multi-view clustering via multi-level contrastive learning," IEEE Trans. on Knowledge and Data Engineering, vol. 37, no. 8, pp. 4716 - 4727, Aug. 2025.
[29] Q. Qi, Z. Qiu, Y. Yan, Y. Lu, and H. Wang, "IMC-Det: Intra-inter modality contrastive learning for video object detection," International Journal of Computer Vision, vol. 133, no. 2, pp. 890-909, 2025.
[30] J. Wen et al., "Dimc-net: Deep incomplete multi-view clustering network," in Proc. of the 28th ACM Int. Conf. on Multimedia, pp. 3753-3761, Seattle, WA, USA, 12-16 Oct. 2020.
[31] C. M. Bishop, Pattern Recognition and Machine Learning, Springer Science+Business Media, 2006.
[32] S. -Y. Li, Y. Jiang, and Z. -H. Zhou, "Partial multi-view clustering," in Proc. of 28th AAAI Conf. on Artificial Intelligence, pp. 1968-1974, Québec, Canada, 27-31 Jul. 2014.
[33] W. Jie, et al., "CDIMC-net: Cognitive deep incomplete multi-view clustering network," in Proc. of the 29th Int. Joint Con. on Artificial Intelligence, pp. 3230-3236, Yokohama, Japan, Jan. 2020.
[34] J. Luo and C. Zhu, "Incomplete multiview clustering with bipartite tensors," Applied Intelligence, vol. 55, no. 13, Article ID: 924, 2025.