一种改进的GA-BP算法在手写体数字识别的应用研究

An Improved GA-BP Algorithm to Handwritten Numeral Recognition Application

在分析GA-BP算法不足的基础上,通过对GA算法中的相应算子进行改进设计,从而有效避免了GA算法中的出现局部次优的情况,并把GA算法产生的最优个体作为BP神经网络的连接权值和阈值,应用于手写体数字识别过程中。实验结果表明,改进的GA-BP算法可以提高BP网络的学习速度和识别效果。

After the analysis of insufficient in the GA-BP algorithm,by GA algorithm in the corresponding operators to improve the design,effectively avoiding GA algorithm in a partial sub-optimal situation and the GA algorithm produced the best individual as the BP neural network connection weights and thresholds,applied to handwritten digit recognition process.Experimental results show that the improved GA-BP algorithm can improve the learning speed of BP network and identification.