摘要
提出了一种基于Q学习的系统服务性能自优化方法.该方法通过感知网络系统的服务性能状态参数变化,利用前馈神经网络的非线性映射关系得到执行动作,综合系统服务性能变化情况及服务的可用性来计算环境奖赏函数值,利用Q学习的自学习特性和预测能力,使系统服务性能达到最优.仿真结果表明,该方法对优化系统整体可信性和服务效用具有明显的优越性.
According to the notion of autonomic computing,a network system service performance online real-time optimization method is proposed based on Q-learning algorithm.First,service performances are taken as parameters,which can affect network system as targets.Second,feed forward neural network is used as nonlinear mapping to gain executive action.Finally,the environment reward function values are calculated according to the change of system service performance and service availability.Then self learning feature and predictive ability of Q-learning is used to make the system service performance achieve optimization.The simulation results show that the optimization method has obvious advantages in credibility and service utility of the overall system maintenance.
出处
《微电子学与计算机》
CSCD
北大核心
2013年第8期63-66,共4页
Microelectronics & Computer
基金
国家自然科学基金项目(61003035
61142002)
河南省科技创新人才计划项目(12410051006)
关键词
自律计算
服务可信性
前馈神经网络
Q学习算法
autonomic computing
service dependability
feed-forward neural network
Q-learning algorithm
