论文部分内容阅读
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。
In this paper, we present a method of single online adjustment of neural network weight and implementation of forecasting based on state-feedback neural network, and put forward the corresponding calculation solution considering that the actual measured values used in forecasting are not easy to obtain. Experiments show that the proposed method combines the network structure obtained by offline learning and online adaptive network adjustment to obtain the ideal real-time prediction result.