Multi-view clustering based on view-attention driven

Multi-view Clustering focuses on discovering coherence information and complementary information about the data among the different views, but often the views are accompanied by related information that is unrelated to the clustering goal. To address this problem, this paper proposes a clustering me...

Full description

Saved in:
Bibliographic Details
Published in:International journal of machine learning and cybernetics 2023-08, Vol.14 (8), p.2621-2631
Main Authors: Ma, Zhifeng, Yu, Junyang, Wang, Longge, Chen, Huazhu, Zhao, Yuxi, He, Xin, Wang, Yingqi, Song, Yalin
Format: Article
Language:English
Subjects:
Algorithms
Artificial Intelligence
Clustering
Complex Systems
Computational Intelligence
Control
Deep learning
Engineering
Mechatronics
Original Article
Pattern Recognition
Representations
Robotics
Systems Biology
Teaching methods
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Multi-view Clustering focuses on discovering coherence information and complementary information about the data among the different views, but often the views are accompanied by related information that is unrelated to the clustering goal. To address this problem, this paper proposes a clustering method named Multi-view Clustering Based on View-Attention Driven. Based on Autoencoder, the method learns feature representations from different views data using contrast learning and attention mechanisms. During the process of learning feature representations of different views, consider information of interest to other views. Guiding information is provided in an attention-driven manner to guide feature learning. On the one hand, it strengthens the focus on information in all views. On the other hand, it reduces the impact on information contained only in a subset of views that isn’t relevant to clustering. In addition, random initialization is used to train the autoencoder, minimizing the network structure’s influence on initialization parameters. Four challenging datasets were used to test the method and it was shown to outperform other competitive multi-view clustering methods. The codes will be available at https://github.com/mzf1998/Multi-view-Clustering-Based-on-View-Attention-Driven .
ISSN:1868-8071
1868-808X
DOI:10.1007/s13042-023-01787-9