论文部分内容阅读
提出一种概率神经网络（ＰＮＮ）的ＥＭ（ＥｘｐｅｃｔａｔｉｏｎＭａｘｉｍｉｚａｔｉｏｎ）训练算法．ＰＮＮ网为一四层前馈网，它构成一个贝叶斯分类器，实现多类分类的贝叶斯判别，它把输入的样本模式，经网络变换为输出的分类判决．其网络节点对应于贝叶斯后验概率公式的各个变量．此ＰＮＮ网络用高斯核的Ｐａｒｚｅｎ窗函数作为核密度函数，网络参数训练由ＥＭ算法实现，其学习方式为类间的监督学习和类内的非监督学习．实验表明了此网络及其学习算法在分类应用中的有效性．
A probabilistic neural network (PNN) EM (ExpectationMaximization) training algorithm is proposed. The PNN network is a four-layer feedforward network that forms a Bayesian classifier that implements Bayesian classification of multiple classes, which transforms input sample patterns into output classification decisions via the network. Its network nodes correspond to the various variables of the Bayesian posterior probability formula. The PNN network uses the Parzen window function of Gaussian kernel as a kernel density function, and the network parameter training is implemented by the EM algorithm. The learning mode is supervised learning between classes and unsupervised learning within a class. Experiments show the effectiveness of this network and its learning algorithm in classification application.