Adversarial Consistency-Guided Representation Learning for Multi-view Clustering

Yuchen Lin; Kunpeng Xu; Ying Fang; Lifei Chen

cs.LGSep 28, 2026

Adversarial Consistency-Guided Representation Learning for Multi-view Clustering

Authors: Yuchen Lin, Kunpeng Xu, Ying Fang, Lifei Chen

Organizations: School of Computer and Cyber Security, Fujian Normal University, Fuzhou 350000, China · Digital Fujian Internet-of-Things Laboratory of Environmental Monitoring, Fujian Normal University, Fuzhou 350000, China · School of Computer Science, McGill University, Montreal H3A 2A7, Quebec, Canada

Abstract

Multi-view clustering aims to capture cross-view consistency while exploiting view-specific information. However, shared representations learned to capture cross-view consistency may still retain view-identifying information, potentially compromising the consistency of cross-view clustering structures. To address this issue, we propose ACGRL, an adversarial consistency-guided representation learning framework for multi-view clustering. ACGRL employs a gradient-reversal view discriminator to reduce view identifiability and obtain invariant reference representations. These representations are then frozen to provide fixed references for disentangling view-specific information from cross-view common information in the subsequent learning stage. The fixed reference representations are concatenated with the learned view-specific representations for reconstruction and clustering, with cross-view cluster alignment encouraging consistent clustering assignments. Experiments on four benchmark datasets demonstrate the superior clustering performance of ACGRL compared with representative multi-view clustering methods.

Figures & tables

Figure. 1 : Overview of ACGRL. Stage I adversarially learns view-invariant representations, which are frozen to provide consistency supervision for view-specific learning in Stage II. The resulting representations are integrated for reconstruction and clustering.

Method	NGs			Hdigit			BBCSport			Cora
Method	ACC	ARI	NMI	ACC	ARI	NMI	ACC	ARI	NMI	ACC	ARI	NMI
MFLVC (CVPR 2022)	49.00	38.28	50.20	99.74	99.42	99.17	64.52	38.77	46.22	46.42	26.31	37.00
DCP (TPAMI 2022)	22.52	0.14	4.64	99.62	99.16	98.78	42.65	5.53	13.96	23.46	-1.18	3.84
MetaViewer (CVPR 2023)	31.30	3.61	17.25	82.35	72.22	77.19	43.24	13.44	17.40	41.07	14.92	21.83
DealMVC (ACM MM 2023)	91.60	79.84	78.51	99.80	99.34	99.33	80.70	60.05	65.59	52.70	27.82	36.96
GCFAggMVC (CVPR 2023)	68.20	44.52	50.94	97.44	94.34	92.96	66.54	40.18	55.71	49.34	23.31	30.98

Table 1: Clustering performance comparison on four multi-view datasets (%). The best results are highlighted in bold with a pink background, while the second-best results are underlined with a gray background.

Dataset	Metric (%)	w/o $\mathcal{L}_{\mathrm{Con}}$	w/o $\mathcal{L}_{\mathrm{Com}}$	w/o $\mathcal{L}_{\mathrm{JRM}}$	w/o $\mathcal{L}_{\mathrm{Clu}}$	Ours
BBCSport	ACC	87.16	92.66	39.45	35.78	93.58
	ARI	71.69	81.95	2.17	0.00	87.34
	NMI	77.59	81.47	6.64	0.00	85.10
Cora	ACC	59.04	58.30	31.37	40.22	64.58
	ARI	32.49	32.02	2.31	11.56	39.58
	NMI	39.42	36.61	4.34	15.36	43.61

Table 2 : Comparison of different loss-function combinations on the clustering task.

Panel of Figure 2 — Figure 2 : Ablation study on different components.

Panel of Figure 3 — Figure 3 : Disentanglement comparison using CCSI.

Panel of Figure 4 — Figure 4 : t-SNE visualizations of the raw features and learned clustering representations on the Cora and Hdigit datasets.

Explore similar work

Learning Disentangled Representations for Generalized Multi-view Clustering
May 15, 2026Xin Zou, Ruimeng Liu, Chang Tang +4Multi-View Adversarial Training
School of Computing and Information Technology, University of Wollongong, NSW, 2500, Australia · School of Software Engineering, Huazhong University of Science and Technology, Wuhan 430074, China
DAS-PMVC: A Framework for Partial Multi-View Clustering via Dual Alignment and Structure Enhancement
Jul 30, 2026Shubin Ma, Liang Zhao, Chuanye He +4Multi-View Multimodal Alignment
Dalian University of Technology Dalian, Liaoning, China · Inspur Group Co., Ltd. Jinan, Shandong, China · China University of Mining and Technology Xuzhou, Jiangsu, China +1
Multi-View Fair Clustering Guided by Cross-View Sensitive Information Discrepancy
Sep 22, 2026Mudi Jiang, Jiahui Zhou, Xinying Liu +2Algorithmic Fairness
School of Software, Dalian University of Technology, Dalian, Liaoning, China

May 15, 2026, cs.CVJ/K move · Enter open · S save

Learning Disentangled Representations for Generalized Multi-view Clustering

Xin Zou, Ruimeng Liu, Chang Tang, Zhenglai Li+3

School of Computing and Information Technology, University of Wollongong, NSW, 2500, Australia · School of Software Engineering, Huazhong University of Science and Technology, Wuhan 430074, China

Multi-View Adversarial Training Complementarity

Open paper AskPDF arXiv

Multi-View Clustering (MVC) has gained significant attention for its ability to leverage complementary information across diverse views. However, existing deep MVC methods often struggle with view-distribution entanglement during cross-view fusion, which hampers the quality of the shared latent space and leads to suboptimal Figures. To address this issue, we propose the Generalized Multi-view Auto-Encoder (GMAE), a framework designed to preserve cross-view complementarity through disentangled representation learning. Specifically, GMAE employs dual-path autoencoders to decouple source features into view-specific and view-common embeddings, facilitating the discovery of clearer clustering structures. We further construct cross-view adversarial discriminators to guide view-specific encoders in capturing more discriminative features. By strategically modulating mutual information, GMAE effectively aligns distributions and prevents representation collapse, ensuring the generation of robust, non-trivial embeddings. Comprehensive experiments on 13 benchmark datasets demonstrate that GMAE consistently outperforms state-of-the-art methods in both complete and incomplete MVC tasks. Our code implementation is available at the repository: https://github.com/obananas/GMAE.

Abstract

Explore similar work

Learning Disentangled Representations for Generalized Multi-view Clustering

DAS-PMVC: A Framework for Partial Multi-View Clustering via Dual Alignment and Structure Enhancement

Multi-View Fair Clustering Guided by Cross-View Sensitive Information Discrepancy