GA-BP混合训练算法应用于心电图自动识别研究

Recognizing ECG by algorithms of BP neural network

利用神经网络进行心电图识别时,存在神经网络网络结构、初始权值以及网络的动量因子、学习参数难以确定,易陷入局部极小、过拟合等问题。遗传算法具有很强的全局寻优能力,能以较大的概率找到全局最优解,提出一种改进的GA-BP混合训练算法,优化神经网络的权值和结构,应用于自动识别心电图,收到良好的效果。

In recognizing ECG by algorithms of BP neural network,the initial weights,network structure and network learning parameters of the neural network contribute to the problems of momentum factor which is dif icult to be determined,local minimum and over-fit ing.As the genetic algorithm is featured with strong global optimization ability and can find global optimal solution with high probability,this paper proposes a hybrid GA-BP training algorithm which can be applied to optimize the structure and weights of the neural network to achieved good ef ects in the automatic recognition of ECG.