TY - JOUR
T1 - Joint Multi-layer Anchor Graph Fusion and Cluster-aware Weight Learning for Multi-view Clustering
AU - Xin, Haonan
AU - Wu, Danyang
AU - Lu, Jitao
AU - Xiao, Jiacong
AU - Li, Huijin
AU - Wang, Rong
AU - Nie, Feiping
N1 - Publisher Copyright:
© 1992-2012 IEEE.
PY - 2026
Y1 - 2026
N2 - Existing multi-layer anchor graph-based multi-view clustering models extend single-layer anchor graph clustering models, and are capable of capturing richer structural information for clustering tasks. However, they face several challenges: 1) existing methods fail to adaptively fuse multi-layer anchor graphs constructed with different numbers of anchors; 2) these models lack sufficient characterization of the local properties of multi-layer anchor graphs and clusters. To address these challenges, we propose a novel Joint Multi-layer Anchor Graph Fusion and Cluster-aware Weight Learning for Multi-view Clustering (MAGF-CWL) model, which aims to fuse multi-layer anchor graphs with varying anchor numbers and leverage the local relationships between anchor graphs and clusters to enhance clustering performance. Specifically, a spatial representation of the anchor graph, rather than the anchor graph itself, is used to construct the fully connected graph subsequently leveraged for graph cut. The cluster-aware weight learning method is introduced to capture the local interactions between graphs and clusters more effectively. To solve the optimization problem in the MAGF-CWL model, we develop an efficient alternating iterative optimization algorithm. Extensive experimental results demonstrate the effectiveness of the proposed method.
AB - Existing multi-layer anchor graph-based multi-view clustering models extend single-layer anchor graph clustering models, and are capable of capturing richer structural information for clustering tasks. However, they face several challenges: 1) existing methods fail to adaptively fuse multi-layer anchor graphs constructed with different numbers of anchors; 2) these models lack sufficient characterization of the local properties of multi-layer anchor graphs and clusters. To address these challenges, we propose a novel Joint Multi-layer Anchor Graph Fusion and Cluster-aware Weight Learning for Multi-view Clustering (MAGF-CWL) model, which aims to fuse multi-layer anchor graphs with varying anchor numbers and leverage the local relationships between anchor graphs and clusters to enhance clustering performance. Specifically, a spatial representation of the anchor graph, rather than the anchor graph itself, is used to construct the fully connected graph subsequently leveraged for graph cut. The cluster-aware weight learning method is introduced to capture the local interactions between graphs and clusters more effectively. To solve the optimization problem in the MAGF-CWL model, we develop an efficient alternating iterative optimization algorithm. Extensive experimental results demonstrate the effectiveness of the proposed method.
KW - Cluster-aware Weight Learning
KW - Multi-layer Anchor Graphs
KW - Multi-view Clustering
UR - https://www.scopus.com/pages/publications/105054013960
U2 - 10.1109/TIP.2026.3741660
DO - 10.1109/TIP.2026.3741660
M3 - 文章
AN - SCOPUS:105054013960
SN - 1057-7149
JO - IEEE Transactions on Image Processing
JF - IEEE Transactions on Image Processing
ER -