Pure kernel graph fusion tensor subspace clustering under non-negative matrix factorization framework

Shuai Zhao, Yilu Zheng, Jinghao Li, Xiaoqian Zhang, Chunlin Tang, Zhen Tan

Abstract: Highlights•This paper proposes a multi-kernel subspace clustering algorithm — PKGT.•We constructed pure local affinity feature graphs for each base kernel matrix.•PKGT effectively avoids interference from kernel noise during the optimization process.•Constructing graph fusion tensor to capture higher-order correlations between kernel data.•Numerous experimental results show that PKGT outperforms existing methods.