Adversarial Consistency-Guided Representation Learning for Multi-view Clustering

2026-09-28

AI Summary

This study addresses the issue in multi-view clustering where shared representations retain view-identifying information, leading to inconsistent cross-view clustering structures. To overcome this, we propose an Adversarial Consistency-Guided Representation Learning (ACGRL) framework. ACGRL introduces a gradient reversal-based view discriminator to eliminate view identifiability and learn invariant reference representations. Furthermore, it pioneers a novel paradigm that freezes these reference representations to guide feature disentanglement, precisely separating view-shared from view-specific information. Combined with cross-view cluster alignment, this approach achieves consistent clustering. Extensive experiments on four benchmark datasets demonstrate that ACGRL significantly outperforms existing representative methods in clustering performance.

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.

Problem

• Multi-view Clustering
• Cross-view Consistency
• View-identifying Information
• Representation Learning

Innovation

• Multi-view Clustering
• Adversarial Consistency
• Gradient Reversal
• Representation Disentanglement
• Cluster Alignment