摘要
针对输入和输出均为时变函数或过程的实际系统建模和仿真问题,提出一种输入和输出均为时变函数的反馈过程神经网络模型.该模型的第1隐层对来自输入层的时变信号进行空间加权聚合和激励运算,并在将其输出传送至第2隐层的同时反馈至输入层;第2隐层完成对其时变输入的空间加权聚合、时间累积聚合和激励运算,并将其输出传送至输出层.给出了相应的学习算法,并以实例验证了该模型及其学习算法的有效性.
in order to model and simulate systems with time-varying functions or processes, a feedback process neural network model with time-varying input and output functions is proposed. The first hidden layer of the model is utilized to accomplish the weighted space aggregation and the activation operation of the time-varying input signal which is received from the input layer. Its output signal is transferred to the second hidden layer and feeds back to the input layer at the same time. The second hidden layer is utilized to accomplish the weighted space aggregation, the time accumulation aggregation and the activation operation of the time-varying output signal from the first hidden layer. Its output signal is transferred to the output layer. The corresponding learning algorithm is developed. The practical example shows the effectiveness of the proposed model and its learning algorithm.
出处
《控制与决策》
EI
CSCD
北大核心
2007年第1期91-94,99,共5页
Control and Decision
基金
国家自然科学基金项目(60373102
60572174)
关键词
反馈过程神经网络
时变函数
航空发动机状态监控
学习算法
时间序列预测
Feedback process neural network
Time-varying function Aircraft engine condition monitoring
Learning algorithm Time series prediction
