模糊超球神经网络的条件重叠学习算法

A Conditional Overlap Learning Algorithm of the Fuzzy Hyperspheres Neural Networks

模糊超球神经网络的聚类学习算法采用形状因子θ实行有指导的学习。针对θ的引入所带来的问题,提出一种新的无指导学习算法——条件重叠学习算法。算法不受模糊超球形状因子θ的影响,学习速度快,学习后的模糊超球个数更少,识别正确率更高。

The class-aggregating algorithm of the fuzzy hypersphere neural networks use a form factor θ to realize a supervised learning. Appointing to the problems produced by θ, a new unsupervised learning algorithm is developed. The new algorithm has nothing to do with the form factor, achieve better classification accuracy with less training time and a fewer number of huperspheres.