学 术 报 告
题目：Spectral clustering based on local similarity measure of shared neighbors
报告人：汪祥 教授 单位：南昌大学数学与计算机学院
时间：2022年4月22日（周五）上午10:00-11:00
#腾讯会议：164-829-694
报告人及报告内容摘要：
汪祥，南昌大学数学与计算机学院教授、博士生导师。江西省新世纪百千万人才工程人选，江西省杰出青年基金获得者，江西省青年科学家（井冈之星）人选，南昌大学香樟英才特聘教授，宝钢全国优秀教师奖获得者，中国工业与应用数学学会理事，中国计算数学学会理事，中国高等教育学会数学专委会常务理事。主要从事数值代数研究，在大规模稀疏线性方程组、大规模稀疏特征值问题、线性和非线性矩阵方程的数值求解等方面取得了一些成果。目前主持(含完成)国家自然科学基金3项及省部级项目十几项。近几年以第一作者或通讯作者在JSC、NLAA、CCP、JCAM等权威期刊上共发表学术论文50多篇。以第一完成人身份获江西省自然科学三等奖1项和江西省教学成果奖二等奖3项。
Abstract: Spectral clustering has become one of typical and effiffifficient clustering methods and has a variety of applications. The critical step of spectral clustering is similarity measurement, which largely deter[1]mines the performance of spectral clustering. In this paper, we propose a novel spectral clustering algorithm based on local similarity measure of shared neighbors. This similarity measurement exploits the local density information between data points based on the weight of shared neighbors
in directed k-nearest neighbor graph and requires only one parameter k, i.e., the number of nearest neighbors. Experimental analysis on eleven synthetic and real world data sets demonstrates that our proposed algorithm outperforms other six existing spectral clustering algorithms in terms of clustering performance such as normalized mutual information.