MCoCo: Multi-level Consistency Collaborative Multi-view Clustering

• DOI: 10.48550/arxiv.2302.13339
• Publication Date: 2023-01-01
• Authors: Yiyang Zhou; Qinghai Zheng; W. B. Yan; Yifei Wang; Jihua Zhu
• Keywords: Feature (linguistics); Feature vector; Representation; Feature Learning

Multi-view clustering can explore consistent information from different views to guide clustering. Most existing works focus on pursuing shallow consistency in the feature space and integrating of multiple into a unified representation for These methods did not fully consider semantic space. To address this issue, we proposed novel Multi-level Consistency Collaborative learning framework (MCoCo) multi-view Specifically, MCoCo jointly learns cluster assignments aligns labels by contrastive learning. Further, designed multi-level collaboration strategy, which utilizes as self-supervised signal collaborate with Thus, levels spaces each other while achieving their own goals, makes mine without fusion. Compared state-of-the-art methods, extensive experiments demonstrate effectiveness superiority our method.