【摘要】 无监督学习矢量量化(LVQ)是一类基于最小化风险函数的聚类方法,文中通过对无监督LVQ风险函数的研究,提出了无监督LVQ算法的广义形式,在此基础上将当前典型的LVQ算法表示为基于不同尺度函数的LVQ算法,极大地方便了学习矢量量化神经网络的推广与应用。通过对无监督LVQ神经网络的改造,得到了基于无监督聚类算法的有监督LVQ神经网络,并将其应用于说话人辨认,取得了满意的结果并比较了几种典型聚类算法的优劣。更多还原

【Abstract】 Unsupervised learning vector quantization is a kind of clustering algorithm which is derived by minimizing the loss function.This paper presents a generalized formulation of unsupervised LVQ,and transfers classical unsupervised LVQ algorithm into learning vector quantization which is based on typically scaling function,its formulation is very convenient for its extension and its application.By unsupervised LVQ neural network improved we can get supervised LVQ neural network which is based on unsupervised LVQ algorithm and apply it to speaker identification,the efficiency of the algoriothms is illustrated by its application and compare them each other.更多还原