摘要
介绍了Levenberg-Marquardt算法来加速神经网络的训练过程,并且为了使得网络回归分析结果具有良好的泛化能力,在训练算法的目标函数中综合了网络权值因素。最后对所给出的算法进行了实例仿真,仿真结果表明该算法不仅具有较好的数据拟合精度,而且具有很好的泛化性能。
This paper introduces Levenberg-Marquardt algorithm to accelerate the training process of artificial neural network. And it also synthesizes the factor of weights in its target function during the training in order to enhance its generalization ability. Finally a simulation about this method was carried out for an instance. The results demonstrate that this method not only has high precision about the training data, but also has good generalization performance.
出处
《计算技术与自动化》
2004年第4期23-24,33,共3页
Computing Technology and Automation
