摘要
该文给出了一种基于状态回归神经网络的单次在线调整神经网络权重和实施预测的方法,同时考虑到用于预测的历史的实际测量值不易获得,而提出了相应计算解决方案。实验证明,所述预测方法将离线学习所得的网络结构与在线自适应网络调整结合起来,能够得到理想的实时预测结果。
A method with sing le time on-line adjusting neural network weights and prediction based on the rec urrent neural network is presented in this paper.In respect to the problem of hardness to obtain the history measured values used for prediction,a relative c alculating approach is proposed to solve this problem.Experiments prove that th e prediction method combines the structure of networks formed in off-line learn ing with that of the on-line adaptive adjusting to access the desired real time predicting results.
出处
《计算机工程与应用》
CSCD
北大核心
2002年第11期45-47,共3页
Computer Engineering and Applications
基金
国家自然科学基金资助项目(编号:69783008)
广东省自然科学基金资助项目(编号:970461)