论文部分内容阅读
Network tomography receives considerable attention in recent years, where link level delay distribution and loss rate are the two characteristics that draw the most of the attention and are treated as separated issues.In this paper, we propose a novel approach that can be used to estimate either of the characteristics or both at the same time.The proposed approach uses a dynamic-programming-like strategy to control the execution of the EM algorithm that significantly reduces the compu tation complexity.Because of this, the proposed approach is scalable.In contrast to those algorithms that trade the scalability for a local or pseudo MLE, the pro posed approach is a global MLE.In addition, the proposed approach can handle the temporal dependence between links.More importantly, the proposed approach is generic that make it not only applicable in network tomography, but also similar problems in other research areas, such as evolution trees or family trees in statistical genetics.