Structural Multi-view Clustering Network via Heterogeneous Random Walks

Ruina Bai, Weifang Ding, Ruizhang Huang, Yongbin Qin, Yanping Chen

Keywords: multi-view clustering; sample structure; heterogeneous graph; heterogeneous random walks

Abstract:

Multi-view clustering, which aims to partition data samples into disjoint clusters by leveraging information from multiple views, has been shown to be highly effective when incorporating structure information, as widely acknowledged in recent works. This paper presents a novel structural multi-view clustering network via heterogeneous random walks, guided by a unified sample-level structure to enhance clustering performance. We first construct a multi-view heterogeneous graph consisting of sample nodes and view nodes, capturing correlations between views while preserving their specific structures. Then, a multi-step random walk strategy on the heterogeneous graph is introduced to explore high-order sample structures across various views, ensuring that each view structure is taken into account. Based on this, a lightweight network is designed to facilitate structure learning both within-view and cross-view, guided by the unified structure derived from heterogeneous random walks, ultimately achieving representations that are conducive to clustering. Extensive experiments on five real-world datasets demonstrate the superiority of the proposed method.

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 9963