Decentralizing Network Inference Problems with Multiple-Description Fusion Estimation (MDFE)英语教案.pdf

2013 Proceedings IEEE INFOCOM Decentralizing Network Inference Problems with Multiple-Description Fusion Estimation (MDFE) Mehdi Malboubi, Cuong Vu, Chen-Nee Chuah Puneet Sharma Dept. of Electrical Computer Engineering Hewlet-Packard Laboratories (HP Labs) University of California, Davis, USA. Palo Alto, California, USA. {mmalboubi, cnvu, chuah}@ puneet.sharma@ Abstract—Two forms of network inference (or tomography) using underlined deterministic or statistical models [3], [4], problems have been studied rigorously: (a) traffic matrix esti- [7], [9]. mation or completion based on link-level traffic measurements, and (b) link-level loss or delay inference based on end-to- Although the uniqueness and accuracy of the solution are end measurements. These problems are often posed as under- important, many network inference problems need to be solved determined linear inverse (UDLI) problems and solved in a in a timely manner for practical deployment. Nevertheless, centralized manner, where all the measurements are collected at a most existing studies attempt to solve the network inference central node, which then applies a variety of inference techniques problem in a one-shot, centralized manner, where all mea- to estimate the attributes of interest. surements are collected at a central node, which then applies This paper proposes a novel framework