基于聚类法的协同神经网络学习算法

根据协同学理论的基本观点（模式识别的过程即为模式形成的过程），对构造出的协同神经网络在模式识别中的应用进行了研究．发现伴随向量的性能直接影响到模式识别的成功率，而伴随向量是由原型向量计算得到．所以原型向量的选择对识别结果有着十分重要的作用．提出了一种基于聚类算法的选择原型向量的方法．通过对近千个样本进行的模拟实验，结果证明这种基于聚类算法的原型向量选择方法很有效，使识别率有了较大的提高．

Learning Algorithm of Synergetic Neural Network Based on Cluster Algorithm

This paper studies application of synergetic neural network in pattern recognition, and finds that adjoin vectors influence the ratio of recognition directly. According to the theory of synergetics, the adjoin vectors are obtained from the prototype vectors, which are very important in pattern recognition. A learning algorithm based on cluster algorithm is presented and the ratio of recognition gets improved a lot.