Consensus Partition Guided Incomplete Multi-View Clustering

Chunyu Yang

doi:10.1109/ACCESS.2025.3543099

Consensus Partition Guided Incomplete Multi-View Clustering

2025-01-01

0

OA

AI

DOI：10.1109/ACCESS.2025.3543099

Original

Original request for help

Save

Abstract

En 中文

Incomplete multi-view clustering aims to derive a comprehensive consensus partition that is shared across all views, a topic that has attracted significant attention in recent years. Most existing methods for incomplete multi-view clustering attempt to directly learn the consensus partition matrix from noisy, complete or incomplete graphs. However, these approaches have limitations: they often overlook the discriminative information within view-specific partition matrices and fail to fully consider the relationships between the partition and the graph. To address these issues, we introduce two new regularization strategies in this paper. One regularity employs a consensus partition matrix to guide the learning process for multiple complete graphs. The other regularity combines multiple view-specific partition matrices to ensure consistency across different views at the discriminative partition level. Building on these concepts, we propose a novel approach for incomplete multi-view clustering, called Consensus Partition-guided Incomplete Multi-view Clustering (CPIMC). This method leverages consensus information at the partition level to complete multiple incomplete graphs and enforce consistency across views. Extensive experiments conducted on several commonly used incomplete multi-view datasets show that CPIMC outperforms state-of-the-art methods in incomplete multi-view clustering.

Keywords:

Manifolds

Matrix decomposition

Tensors

Laplace equations

Solid modeling

Optimization

Indexes

Hands

Coherence

Stacking

Incomplete multi-view clustering

consensus partition

complete graphs learning

discriminative information

Consensus Partition Guided Incomplete Multi-View Clustering

Abstract

Journal

Organization