Auto-Weighted Multi-View One-Step Spectral Clustering Based on Multi-Graph Fusion

Abstract

Multi-view spectral clustering provides powerful modeling tools for clustering tasks in complex scenarios by fusing the complementary information of multi-source data. However, the existing methods have the following three defects: (1) failing to fully exploit the complementarity and differences among views, leading to insufficient modeling of cross-view consistency; (2) being sensitive to non-Gaussian noise, causing the learned affinity matrices to fail to satisfy the block-diagonal property required for clustering and resulting in poor clustering performance; and (3) being unable to dynamically reflect the true importance of views, thereby restricting the model’s generalizability. To address these issues, we propose an Auto-weighted Multi-view One-step Spectral Clustering method Based on Multi-graph Fusion. The proposed method unifies view-specific affinity matrix learning, consensus affinity matrix learning, adaptive view weighting, and clustering into a single framework. Experimental results on six public datasets illustrate the effectiveness of the proposed model.