A saddle point-guided clustering algorithm for data with complex structure

doi:10.1016/j.knosys.2025.114726

A saddle point-guided clustering algorithm for data with complex structure

2025-10-28

0

PRE

AI

DOI：10.1016/j.knosys.2025.114726

Original

Original request for help

Save

Abstract

En 中文

• We present a novel density-based clustering algorithm that combines a mode-seeking phase with a fragmented cluster fusion phase. • A new method for forming initial clusters is proposed, which is based on the k-mutual nearest neighbors consistency principle and a majority voting mechanism. • A saddle point-guided cluster similarity is designed to measure the homogeneity of two initial clusters. • The proposed algorithm has the ability to discover clusters of different densities and sizes. • Experiments on synthetic and real datasets demonstrate the superiority of the proposed algorithm.

A saddle point-guided clustering algorithm for data with complex structure

Abstract

Journal

Organization