有噪网络断层扫描方法研究  

RESEARCH ON NOISY NETWORK TOMOGRAPHY METHOD

作　　者：吴辰文 朱建东 闫光辉 郑恒 张烨 Wu Chenwen;Zhu Jiandong;Yan Guanghui;Zheng Heng;Zhang Y e(School of Electronic and Information Engineering , Lanzhou Jiaotiong University, Lanzhou 7 3 0 0 7 0 , Gansu, China)

机构地区：[1]兰州突道大学电子与狺息工程学脘,甘肃兰州730070

出　　处：《计算机应用与软件》2016年第8期150-152,205,共4页Computer Applications and Software

基　　金：国家自然科学基金项目(61163010);兰州市科技计划基金项目(2009-1-5);甘肃省自然科学基金项目(1308RJZA111)

摘　　要：噪声数据在一定程度上影响了网络断层扫描的准确性。针对之前网络断层扫描方法大都忽略噪声影响的不足,提出SAK算法。基于卡茨马尔兹算法和SA算法的SAK算法更具有一般性和实时性,SAK算法模仿了原始Kaczmarz算法的特性。实验结果显示,通过用SAK算法处理估计的初始值,使其估计值能够收敛到真实值,在很大程度上能达到去除噪声的目的。Noisy data affects the accuracy of network tomography to some extent. In light of the insufficiency of previous network tomographymethods that they mostly ignore the influence of noise,we proposed SAK algorithm. The SAK algorithm is based on Kaczmarz algorithmand SA algorithm, and is of more universal and real-time; SAK algorithm simulates the characteristics of original Kaczmarz algorithm. Experimentalresults showed that by using SAK algorithm to deal with the estimated initial values, they could converge to the real one; to a great extentit was able to achieve the purpose of noise removal.

关 键 词：网络断层扫描 随机逼近算法 KacarmK算法 SAK算法 网络测量 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]