Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Trusted Mamba Contrastive Network for Multi-View Clustering

Jian Zhu, Xin Zou, Lei Liu, Zhangmin Huang, Ying Zhang, Chang Tang, Li-Rong Dai

TL;DR

The paper tackles untrusted fusion in deep multi-view clustering by introducing the Trusted Mamba Contrastive Network (TMCN), which combines a selective Fusion module (TMFN) with an intra-cluster contrastive objective (AsCL). TMFN performs noise-robust fusion through a Mamba-based selective mechanism, while AsCL aligns view representations within the same cluster to support clustering objectives. On four public multi-view datasets, TMCN achieves state-of-the-art performance across ACC, NMI, and PUR, with ablations confirming the individual contributions of TMFN and AsCL. This approach advances robust, unsupervised multi-view clustering by addressing both view-level noise and cluster-level representation alignment, with public code available.

Abstract

Multi-view clustering can partition data samples into their categories by learning a consensus representation in an unsupervised way and has received more and more attention in recent years. However, there is an untrusted fusion problem. The reasons for this problem are as follows: 1) The current methods ignore the presence of noise or redundant information in the view; 2) The similarity of contrastive learning comes from the same sample rather than the same cluster in deep multi-view clustering. It causes multi-view fusion in the wrong direction. This paper proposes a novel multi-view clustering network to address this problem, termed as Trusted Mamba Contrastive Network (TMCN). Specifically, we present a new Trusted Mamba Fusion Network (TMFN), which achieves a trusted fusion of multi-view data through a selective mechanism. Moreover, we align the fused representation and the view-specific representation using the Average-similarity Contrastive Learning (AsCL) module. AsCL increases the similarity of view presentation from the same cluster, not merely from the same sample. Extensive experiments show that the proposed method achieves state-of-the-art results in deep multi-view clustering tasks. The source code is available at https://github.com/HackerHyper/TMCN.

Trusted Mamba Contrastive Network for Multi-View Clustering

TL;DR

The paper tackles untrusted fusion in deep multi-view clustering by introducing the Trusted Mamba Contrastive Network (TMCN), which combines a selective Fusion module (TMFN) with an intra-cluster contrastive objective (AsCL). TMFN performs noise-robust fusion through a Mamba-based selective mechanism, while AsCL aligns view representations within the same cluster to support clustering objectives. On four public multi-view datasets, TMCN achieves state-of-the-art performance across ACC, NMI, and PUR, with ablations confirming the individual contributions of TMFN and AsCL. This approach advances robust, unsupervised multi-view clustering by addressing both view-level noise and cluster-level representation alignment, with public code available.

Abstract

Multi-view clustering can partition data samples into their categories by learning a consensus representation in an unsupervised way and has received more and more attention in recent years. However, there is an untrusted fusion problem. The reasons for this problem are as follows: 1) The current methods ignore the presence of noise or redundant information in the view; 2) The similarity of contrastive learning comes from the same sample rather than the same cluster in deep multi-view clustering. It causes multi-view fusion in the wrong direction. This paper proposes a novel multi-view clustering network to address this problem, termed as Trusted Mamba Contrastive Network (TMCN). Specifically, we present a new Trusted Mamba Fusion Network (TMFN), which achieves a trusted fusion of multi-view data through a selective mechanism. Moreover, we align the fused representation and the view-specific representation using the Average-similarity Contrastive Learning (AsCL) module. AsCL increases the similarity of view presentation from the same cluster, not merely from the same sample. Extensive experiments show that the proposed method achieves state-of-the-art results in deep multi-view clustering tasks. The source code is available at https://github.com/HackerHyper/TMCN.

Paper Structure

This paper contains 12 sections, 18 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. The Proposed Methodology
  3. Multi-view Data Reconstruction
  4. Trusted Mamba Fusion Network
  5. Average-similarity Contrastive Learning
  6. Clustering module
  7. Experiments
  8. Experimental Settings
  9. Experimental comparative results
  10. Ablation Study
  11. Convergence, Visualization, and Parameter Analysis.
  12. Conclusion and Future work

Figures (3)

  • Figure 1: Overall Framework of TMCN. The framework consists of TMFN and AsCL. TMFN segments the one-dimensional feature vector of each view into sequence vectors, followed by a trusted fusion of multi-view features via a selective mechanism of the Mamba network. In contrast, AsCL is introduced to enhance the similarity of view representations within the same cluster, rather than merely focusing on the similarity at the individual sample level. It further improves the trusted fusion of multi-view data.
  • Figure 2: The convergence analysis and visualization analysis on Hdigit.
  • Figure 3: The parameter analysis on Hdigit.