摘要
共享近邻聚类(SNN)是一种基于图的聚类算法,能够在不预设聚类数目的前提下,很好得区分彼此相似的邻近簇。然而SNN因计算开销太大,不适于处理大数据量、高属性维数据。P系统是一种并行分布式生物计算模型,具有与图灵机等价的强大计算能力。本文将SNN与P系统相结合,设计了一种含有多促进剂和多抑制剂的类细胞P系统,提出了基于该系统的膜聚类算法,称为共享近邻膜聚类算法(SNN-P)。最后,用Olivetti Face数据集验证了SNN-P在人脸识别中的有效性。理论分析表明SNN-P相比于传统聚类算法具有极低的时间复杂度,实验结果表明SNN-P对面部图像具有良好的识别能力。
Shared Nearest Neighbor( SNN) is a clustering algorithm based on graph theory,which can distinguish similar neighboring clusters without presetting cluster numbers. However,SNN is not suitable for processing large data sets and high dimensional data sets due to its high computational cost. As a powerful distributed parallel computing device,P system has the computing power equivalent to Turing machine. In this paper,SNN and P system are combined to design a cell-like P systems with multi-promoters and multi-inhibitors. Based on this system,a novel membrane clustering algorithm named Shared Nearest Neighbor Clustering Algorithm Based on P System( SNN-P) is proposed.In addition,the Olivetti Face dataset is used to illustrate the effectiveness of SNN-P in face recognition. The theoretical analysis shows that SNN-P has a very low time complexity compared with traditional clustering algorithm. Experimental results show that SNN-P has a good ability in facial image recognition.
作者
王鑫
刘希玉
顾飞
WANG Xin;LIU Xiyu;GU Fei(Business School, Shandong Normal University, Jinan 250014, China)
出处
《激光杂志》
北大核心
2019年第3期74-78,共5页
Laser Journal
基金
国家自然科学基金项目(No.61876101
No.61806114
No.61602285
No.61602284
No.61502283
No.61472231)
关键词
膜计算
P系统
共享近邻
聚类分析
图像处理
membrane computing
P systems
shared nearest neighbor
clustering analysis
image processing