基于NT技术的链路时延估计研究通常使用极大似然估计方法,EM算法是极大似然估计的有效算法,特别适合非完全数据的参数估计,但是EM算法的计算量会随着网络规模的增长变得相对复杂,而导致收敛速度慢和容易局部最大化等问题,从而影响在网络中的应用.对此提出了一种将网络拓扑进行2步骤划分的方法,对2步划分出的局部拓扑分别进行EM算法的极大似然估计,降低原有方法的计算复杂度.最后利用NS2仿真实验验证了该算法的可行性和准确性.%The research of link delay estimation algorithm usually adopts the maximum likelihood estimation algorithm.EM algorithm is an effective algorithm of the maximum likelihood estimation algorithm,and it is appropriate for parameter estimation of incomplete data.But with the increase of network scale,the computation of EM algorithm may be more complex,and it will lead to a slow convergence speed and a local maximization problem,which affects the application in the network.This paper proposes a method that can divide the network topology by two steps,and then have a estimation of EM algorithm to each divided topology.This method can reduce the computational complexity of the original method.Finally,the feasibility and accuracy of this method are verified.
展开▼