RAC-DMVC: Reliability-Aware Contrastive Deep Multi-View Clustering under Multi-Source Noise

TL;DR

This work tackles reliable multi-view clustering under multi-source noise by introducing RAC-DMVC, which combines a reliability graph-guided noise contrastive objective, cross-view reconstruction for data-level denoising, dual-attention imputation for missing views, and a self-supervised cluster distillation mechanism. The approach jointly addresses missingness and observation noise, yielding robust representations and improved clustering performance across five benchmarks with varying noise ratios. Empirical results demonstrate state-of-the-art performance and resilience to noise, supported by ablations and visualizations that confirm each module's contribution. The framework offers practical benefits for real-world, noisy multi-view data applications by maintaining stable clustering quality despite significant data degradation.

Abstract

Multi-view clustering (MVC), which aims to separate the multi-view data into distinct clusters in an unsupervised manner, is a fundamental yet challenging task. To enhance its applicability in real-world scenarios, this paper addresses a more challenging task: MVC under multi-source noises, including missing noise and observation noise. To this end, we propose a novel framework, Reliability-Aware Contrastive Deep Multi-View Clustering (RAC-DMVC), which constructs a reliability graph to guide robust representation learning under noisy environments. Specifically, to address observation noise, we introduce a cross-view reconstruction to enhances robustness at the data level, and a reliability-aware noise contrastive learning to mitigates bias in positive and negative pairs selection caused by noisy representations. To handle missing noise, we design a dual-attention imputation to capture shared information across views while preserving view-specific features. In addition, a self-supervised cluster distillation module further refines the learned representations and improves the clustering performance. Extensive experiments on five benchmark datasets demonstrate that RAC-DMVC outperforms SOTA methods on multiple evaluation metrics and maintains excellent performance under varying ratios of noise.

Figures (5)

• Figure 1: Overview of the our pipeline. The processing flow begins with cross-view reconstruction for robust representation. Subsequently, a similarity-based reliability graph is constructed to guide both contrastive learning and a dual-attention imputation module. Finally, self-supervised clustering distillation is applied to refine the view-specific representations.
• Figure 2: Ablation study on five dataset with 50% noise ratio
• Figure 3: Visualization of the training process on Caltech101. 0 and 99 represent the epoch of training.
• Figure 4: The impact of $\sigma$ on performance.
• Figure 5: The impact of $\tau_c,\tau_d$ on performance.