摘要
识别网络内部的故障链路对提升网络性能具有重要参考价值。研究了树型拓扑下基于端到端测量的故障链路诊断问题,提出一种最大伪似然估计方法估计链路先验故障概率,把树型拓扑划分为一系列具有两个叶节点的子树,并使用期望最大化(EM)算法最大化每个子树的似然函数,求出链路先验概率。仿真实验表明,该方法与现有的联立方程组求解方法估计精度相当,但是大大降低了算法时间复杂度,证明了该方法的有效性。
Identifying internal lossy link has a great reference value to enhance the performance of the network.This paper studied the lossy link diagnostic problems of tree topology which based on end-to-end measurement,proposed a method of maximum pseudo likelihood criterion to estimate the prior probability of the link.The method first divided the tree topology into a series of two-leaf-node subtree,and then used the EM algorithm to maximize the likelihood function for each subtree to get the link prior probability.Simulation results show that,compare with current simultaneous equations method,the new method does not reduce the precision,but greatly reduces the time complexity of algorithm,that demonstrate the validity of the method.
出处
《计算机应用研究》
CSCD
北大核心
2012年第4期1514-1517,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(60872033)
关键词
故障链路诊断
端到端测量
最大伪似然估计
期望最大化算法
lossy link identification
end-to-end measurement
maximum pseudo likelihood estimation
EM algorithm
