无限制性手写数字的多神经网络识别

A Combinatoral Structure of Multiple Neural Network for Hand-Written Digits Recognition

研究一种多神经网络的组合结构，这种组合结构有效地克服了ＢＰ算法学习速度慢，难以收敛的主要缺点多神经网络的组合结构应用于手写数字识别。

In this paper, a combinational structure of the multiple neural network is presented It can overcome the BP algorithm′s weakness that learning is slow and convergence is difficult The combinatorial structure of the multiple neural network has been applied successfully to hand-written digits recognition Evident effects are obtained