基于免疫聚类的RBF神经网络研究

Study on modified RBF neural networks based on immune clustering

通过分析以往人工免疫聚类算法的不足之处,提出了一种改进的基于人工免疫聚类与RBF神经网络的混合算法。该算法由两个阶段组成:第一阶段采用人工免疫机制来确定RBF网络隐层的聚类中心的位置和数量;第二阶段建立RBF神经网络,对输入样本数据进行学习、训练,求输出层的权值矩阵。最后以肝病病证诊断进行仿真,建立基于免疫聚类的RBF网络模型。实验结果表明:该算法用于中医病证诊断的研究是可行的和有效的。

A modified hybrid algorithm for artificial immune clustering and RBF neural networks is presented, the weaknesses of the former artificial immune clustering algorithm is analyzed. The algorithm includes two parts. The first part is that artificial immune clustering is used to determine the initial positions and number of the clustering centers in the RBF network. The second part is to set up the RBF neural networks to learn and train the input sample data, and to calculate the weight matrix W. Finally, it uses the hepatic disease symptom as simulation. The experimental result demonstrates that the algorithm for TCM diagnose is possible and effective.