This article presents a new method for multiview clustering (MVC) via contrastive learning, called CCEC. MVC is a process that groups data samples into clusters by synthesizing information across multiple views. Although deep learning methods have shown strong capabilities in MVC scenarios, generalizing feature representations while maintaining consistency remains a challenge. The proposed method, CCEC, addresses this by incorporating semantic connection blocks into a feature representation to preserve consistent information among multiple views. The representation process for clustering is enhanced through spectral clustering, and consistency across multiple views is improved. Experimental results show that CCEC outperforms state-of-the-art methods.
Publication date: 24 Jan 2024
Project Page: https://anonymous.4open.science/r/CCEC-E84E/
Paper: https://arxiv.org/pdf/2401.12648
