摘要
利用由多个关键报告生成的插值多项式预测方程来预测未来几个周期的业务量,通过1个消息(grant)对多个周期的授权,节省了带宽和减少了数据包的时延,建立了知识支撑系统(KSS)系统,通过偏导(DC)寻优法来优化关键周期和预测周期的个数,通过神经网络来优化非线形预测方程的调节因子.仿真结果很好地验证了它的效果.
The mechanism forecasts the traffic for many cycles through the prediction function of interpolating polynomial produced by the reports of many key cycles, and grants many cycles at a time. Not only can it save bandwidth, but also reduce the data packet delay. A knowledge support system (KSS) is set up. The key cycle number and predictive cycle number is optimized by the differential coefficient (DC) optimizing method, the adaptive factors of the non - lineal prediction function are optimized by the neural network. Simulation verifies its merits very well.
出处
《北京邮电大学学报》
EI
CAS
CSCD
北大核心
2008年第5期61-64,共4页
Journal of Beijing University of Posts and Telecommunications
基金
国家“973计划”项目(2007CB307101)
国家“863计划”项目(2007AA012203)
关键词
以太无源光网络
神经网络
轮巡
优化
Ethernet passive optical network
neural network
polling
optimize