为减少违反三角形不等式(TIV)对因特网时延空间建模的影响,提出了一种基于空间修复的因特网时延空间嵌入算法S-Vivaldi。S-Vivaldi先对原有的时延空间D进行指数变换修复,获得一个几乎没有TIV的距离矩阵D',再对D'采用Vivaldi算法进行坐标嵌入,当需要预测任意2个节点间的时延时,再进行一系列的逆变换。实验表明,该算法在大多数情况下可以显著提高预测精度。
To counteract TIV's influence on the Internet delay space models,an Internet delay space embedding algo-rithm based space repairing S-Vivaldi(scaling-Vivaldi) was proposed.S-Vivaldi firstly applied an exponent transforma-tion on the delay matrix D,retrieved a matrix D' almost without TIV,and then embedded D' by the Vivaldi algorithm.The delay between any two nodes could be computed by a series of inverse transformation.Experimental results show that S-Vivaldi could improve the prediction accuracy apparently in the most case.