Incremental Nyström-based Multiple Kernel Clustering

• DOI: 10.1609/aaai.v39i16.33825
• Publication Date: 2025-04-11T12:39:27Z
• Authors: Yu Feng; Weixuan Liang; Xinhang Wan; Jiyuan Liu; Suyuan Liu; Qian Qu; Renxiang Guan; Huiying Xu; Xinwang Liu
• Keywords: Kernel (algebra)

Existing Multiple Kernel Clustering (MKC) algorithms commonly utilize the Nyström method to handle large-scale datasets. However, most of them employ uniform sampling for kernel matrix approximation, hence failing accurately capture underlying data structure, leading large approximation errors. Additionally, they often use same landmark points all approximations, reducing diversity. Moreover, in scenarios where approximate matrices emerge over time, these methods require storing historical information and recalculating, resulting inefficient resource utilization. To address issues, we propose a novel MKC algorithm, termed Incremental Nyström-based (INMKC). Specifically, leverage score is utilized reduce errors enhance Furthermore, consensus clustering structure that aligns with newly emerged base updates, avoiding recalculating previous matrices, thus saving substantial computational resources. tackle challenge aligning incremental kernels different points. Extensive experiments on proposed INMKC demonstrate its effectiveness efficiency compared state-of-the-art methods.