基于神经网络的聚类算法研究  

Research on clustering algorithm based on neural network

作　　者：胡钦华 党涛 邓见光 HU Qinhua;DANG Tao;DENG Jianguang(School of Computer Science&Technology,Dongguan University of Technology,Dongguan 523808,China)

机构地区：[1]东莞理工学院计算机科学与技术学院,广东东莞523808

出　　处：《电子设计工程》2024年第22期186-190,共5页Electronic Design Engineering

基　　金：广东省高校特色创新项目(2018KTSCX221)。

摘　　要：聚类作为一种无监督学习算法,在处理大规模数据、参数值选择、噪声异常值等方面效果不佳。为了充分考虑算法的有效性和稳定性,该文将神经网络与传统聚类算法相结合,介绍了自组织特征映射、学习向量量化和自适应共振理论三种基于神经网络的聚类算法。并结合常见的距离度量和相似性度量来验证该类算法在自适应学习和噪声容忍性方面的优势,从而弥补传统聚类算法的不足。全文工作将进一步指导我们深入研究聚类算法在深度神经网络中的应用。As an unsupervised learning algorithm,clustering is not effective in dealing with large⁃scale data,parameter value selection,noise outliers and so on.In order to fully consider the effectiveness and stability of the algorithm,this paper combines the neural network with the traditional clustering algorithm,and introduces three clustering algorithms based on the neural network:self⁃organizing feature mapping,learning vector quantization and adaptive resonance theory.The advantages of these algorithms in adaptive learning and noise tolerance are verified by using common distance and similarity measures,so as to make up for the shortcomings of traditional clustering algorithms.This paper will further guide us to further study the application of clustering algorithm in deep neural networks.

关 键 词：神经网络 聚类算法 距离度量 相似性度量 

分 类 号：TN-9[电子电信]