Journal of Frontiers of Computer Science and Technology ›› 2022, Vol. 16 ›› Issue (12): 2809-2819.DOI: 10.3778/j.issn.1673-9418.2104019

• Artificial Intelligence • Previous Articles     Next Articles

Nearest Neighbor Label Propagation for Density Peak Clustering

SONG Peng1,2, GE Hongwei1,2,+()   

  1. 1. School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122, China
    2. Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence, Jiangnan Uni-versity, Wuxi, Jiangsu 214122, China
  • Received:2021-04-08 Revised:2021-05-26 Online:2022-12-01 Published:2021-05-18
  • About author:SONG Peng, born in 1996, M.S. candidate, stu-dent member of CCF. His research interests include pattern recognition and machine learning.
    GE Hongwei, born in 1967, Ph.D., professor, Ph.D. supervisor. His research interests include artificial intelligence, pattern recognition, machine learning, image processing and analysis, etc.
  • Supported by:
    National Natural Science Foundation of China(61806006);Priority Academic Development Program of Jiangsu Higher Education Institutions.


宋鹏1,2, 葛洪伟1,2,+()   

  1. 1.江南大学 人工智能与计算机学院,江苏 无锡 214122
    2.江苏省模式识别与计算智能工程实验室(江南大学),江苏 无锡 214122
  • 通讯作者: +E-mail:
  • 作者简介:宋鹏(1996—),男,辽宁丹东人,硕士研究生,CCF学生会员,主要研究方向为模式识别、机器学习。
  • 基金资助:


Dynamic graph-based label propagation for density peaks clustering (DPC-DLP) is an improved algo-rithm of density peaks clustering (DPC). The related parameters involved in the algorithm are too complex and the algorithm uses labeled data in each iteration, which will lead to the expansion of label errors and the deterioration of clustering effect due to too many iterations. To solve the above problems, this paper proposes a nearest neighbor label propagation for density peak clustering (DPC-NLP), which mainly has three steps. Firstly, the local density and the minimum distance are used to score the sample points and the clustering center is determined according to the score. Then, the clustering backbone is formed by the labels of the clustering center in its nearest neighbor. Finally, the label propagation method based on the nearest neighbor is used to propagate the labels of the clustering back-bone to the remaining samples and the final clustering results are formed. The nearest neighbor label propagation algorithm takes full account of the structural association between samples, constantly updates the state of data in the process of propagation, and uses more sufficient information to improve the accuracy of allocation. The algorithm is verified on the synthetic and real-world datasets and compared with the current mainstream clustering algorithms. Ex-perimental results show that DPC-NLP is superior in performance and robustness and can deal with complex data such as manifold and nonlinear data.

Key words: clustering, density peaks clustering, label propagation, nearest neighbor



关键词: 聚类, 密度峰值聚类, 标签传播, 最近邻

CLC Number: