摘要
基于最大似然的网络拓扑估计方法能够获得全局最优的估计结果,优于一般局部最优化和节点对融合方法,但在网络规模较大时存在计算复杂度较高的缺点。首先证明了网络拓扑估计似然函数是单峰的且峰值为最大值;然后利用单峰特征,改进了现有基于最大似然的拓扑估计方法,降低了计算复杂度。最后,Matlab和NS-2仿真结果证明,在不降低拓扑估计准确率的情况下,改进算法将计算复杂度减少了30%～40%。
Maximum likelihood based topology estimation method can obtain globally optimal estimation resuhs, and its performance is better than that of other methods such as general local optimization and node fusion method. However, when network scale is large, the topology estimation is computationally intensive. To solve this problem, firstly, the authors proved that the network topology estimation likelihood function is a single-peaked function and the peak value is maximum value. Using the single-peaked property of likelihood function, the current maximum likelihood based topology estimation method had been improved. Finally, the results of Matlab and NS-2 simulations show that the improved method cuts computational complexity by 30% -40% without reducing topological estimation accuracy.
出处
《计算机应用》
CSCD
北大核心
2011年第1期212-214,238,共4页
journal of Computer Applications
基金
国家自然科学基金资助项目(60872033)
新世纪优秀人才支持计划资助项目(NECT-07-0148)
关键词
拓扑估计
最大似然
网络层析成像
三明治包
马尔可夫链蒙特卡洛算法
topology estimation
maximum likelihood
network tomography
sandwich packet
Markov Chain Monte Carlo （MCMC） algorithm
