摘要
在分析了传统神经网络预测模型的基础上,提出用模拟退火算法代替局部梯度下降法修正网络权值,从而避免了人工神经网络容易陷入局部极小的缺陷.经实例证实,该方法能有效地提高预测的精度.
After analyzing the conventional artificial neural network prediction model,we propose to modify the weight with Simulated Annealing instead of local gradient descend.Thereby, prevent the artifical neural network from resulting in local minimum. This method is able to efficiently improve the prediction accuracy by case.
出处
《喀什师范学院学报》
2004年第3期63-65,共3页
Journal of Kashgar Teachers College
