摘要
本文提出了一种改进的BP算法以优化神经网络连接权 ,使网络具有快速全局收敛的能力 ,由此建立的温度预测模型取得了较好的效果。实验中选取了多组数据对网络进行训练和测试 ,在此过程中 ,对学习率和动态参数的选取以及网络结构的优化进行了初步探讨 。
This paper presents an improved BP algorithm to optimizing weights of neural network,so that the network can be constringed across-the-board and fast.On the basis of it,satisfactory effect was made in the forecasting meteorologic temperature model.Many groups of data were selected to train and test this network,in this procedure,the choice of learning rate & momentum factors,and optimizing the structure of this network were inquired into,and some profitable conclusions were obtained.
出处
《计算机应用与软件》
CSCD
北大核心
2004年第3期108-110,共3页
Computer Applications and Software
