Missing Pattern Tree based Decision Grouping and Ensemble for Deep Incomplete Multi-View Clustering

TL;DR

This paper tackles deep incomplete multi-view clustering under highly inconsistent missing patterns by proposing TreeEIC, which uses a missing-pattern tree to partition data into pattern-consistent decision sets, enabling full utilization of cross-view pairs. Clustering decisions from these sets are fused via an uncertainty-weighted ensemble, and the ensemble knowledge is distilled back to each view through cross-view consistency and inter-cluster discrimination losses. The method is imputation-free and demonstrates state-of-the-art performance and robustness across six benchmark datasets, especially under extreme missingness. The approach offers a practical, scalable solution for robust multi-view learning where view availability is highly variable.

Abstract

Real-world multi-view data usually exhibits highly inconsistent missing patterns which challenges the effectiveness of incomplete multi-view clustering (IMVC). Although existing IMVC methods have made progress from both imputation-based and imputation-free routes, they have overlooked the pair under-utilization issue, i.e., inconsistent missing patterns make the incomplete but available multi-view pairs unable to be fully utilized, thereby limiting the model performance. To address this, we propose a novel missing-pattern tree based IMVC framework entitled TreeEIC. Specifically, to achieve full exploitation of available multi-view pairs, TreeEIC first defines the missing-pattern tree model to group data into multiple decision sets according to different missing patterns, and then performs multi-view clustering within each set. Furthermore, a multi-view decision ensemble module is proposed to aggregate clustering results from all decision sets, which infers uncertainty-based weights to suppress unreliable clustering decisions and produce robust decisions. Finally, an ensemble-to-individual knowledge distillation module transfers the ensemble knowledge to view-specific clustering models, which enables ensemble and individual modules to promote each other by optimizing cross-view consistency and inter-cluster discrimination losses. Extensive experiments on multiple benchmark datasets demonstrate that our TreeEIC achieves state-of-the-art IMVC performance and exhibits superior robustness under highly inconsistent missing patterns.

Figures (7)

• Figure 1: Framework Overview of Our TreeEIC. First, view-specific models generate sample embeddings via autoencoders. (a) The embeddings are grouped into multiple decision subsets according to missing patterns. Samples in each subset share the consistent missing pattern, allowing to obtain their clustering decisions in one feature space. (b) Clustering decisions from all subsets are aligned and weighted based on uncertainty to produce the ensemble robust clustering knowledge. (c) Knowledge distillation is then applied to transfer the ensemble results to individual view-specific models, responsible for cross-view consistency via $\mathcal{L}_{cons}$ and inter-cluster discrimination via $\mathcal{L}_{disc}$.
• Figure 1: ACC $vs.$ Missing Rate across six datasets. When the missing rate $\tau =1.0$, i.e., the highly inconsistent missing patterns, we can observe that most IMVC methods have heavy performance degradation while our method TreeEIC is still robust.
• Figure 2: Illustration of the missing patterns in incomplete multi-view data. Numerous incomplete multi-view data exhibiting inconsistent missing patterns can be grouped into multiple decision sets. The missing pattern of samples in each decision set exhibits consistency and thus the pair relationship in these samples can be exploited to improve IMVC.
• Figure 3: ACC $vs.$ Missing Rate on AWA-7 and ModelNet40. When the missing rate $\tau =1.0$, i.e., the highly inconsistent missing patterns, we can observe most IMVC methods have heavy performance degradation while our method TreeEIC is still robust.
• Figure 4: ACC $vs.$ Loss on HandWritten and Caltech101-7.