本文考慮一樹狀網路模型(Tree-type network model)，此網路模型在現實生活中有廣泛的應用(如無線網路(wireless network)，電話交換中心(call centers)等)。我們所關心的是此樹狀網路中每個佇列的服務品質(Quality of Service; QoS)。但是受限於網路權限(authorization)，通訊法規(protocol)等限制，系統管理者往往無法得到完整的佇列資訊。所以本篇論文主旨在於如何透過可觀測的佇列來估計其他無法觀測的佇列資訊，這一類的問題被稱為網路診斷
(Network Tomography)。本文的方法是藉由建立一統計模型來推估感興趣的參數值，而這種利用可觀測到的有限資訊來推測無法觀測的佇列資訊的統計問題，我們稱之為Statistical Inverse Problem。最後我們透過電腦模擬來評估所提方法對佇列工作遺失率的估計準確性，並提出改善系統表現值的方法。

Assessing and monitoring the performance of computer and communi-
cations networks is an important problem for network engineers. We con-
sider a tree-type network model, and it has a broad application in real life(
wireless network, call centers, etc.). Our focus here is on estimating and monitoring network Quality-of-Service (QoS) parameters. But constrain-
ed by authorization or protocol, network engineers usually can’t get queues information completely. We discuss methods for estimating edge- level parameters from end-to-end path-level measurements, an important engineering problem that raises interesting statistical modeling issues. This kind of question is called “Network Tomography”. The method of this paper is to estimate interested parameters by constructing a statistical model, and we call this statistical issue “Statistical Inverse Problem”. In the end we will show a simulation result and advance a method to improve model performance.

[1] George Michailidis, Bowel Xi, Vijayan N.Nair.
“Statistical aspects of the analysis of data
networks”, Technometrics , 49, pp.318-334, 2007.
[2] West. M., (2000) “Bayesian regression analysis in the “Large p, Small n” paradigm”, Technical Report 00-22, Institute of Statistics and Decision Sciences, Duke University.
[3] Geo5204 “Minimum norm solutions” Inverse theory lecture 6.
http://www.geo.umn.edu/courses/5204/Lectures/lecture_06.pdf
[4] Zellner. A. (1986) “On assessing prior distributions and Bayesian regression analysis with g-prior distribution”, In Bayesian Inference and Decision Techniques: Essays in Honor of Bruno de Finetti, pp233-243.Amsterdam:North-holland.
[5]West.M., Nevins K.R., Marks J.R., Spang,R., Blanchett,C. and Zuzan,H. (2000) “DNA microarray data analysis and regression modeling for genetic expression profiling.” ISDS discussion Paper #00-15. Submitted for publication.
[6] Douglas C. Montgomery, Raymond H. Myers. (2002) “Response surface methodology : process and product in optimization using designed experiments. 2nd edition” New York:Wiley
[7] M. Armony and N. Bambos, (2003) “Queueing Dynamics and Maximal Throughput Scheduling in Switched Processing Systems” , Queueing Systems: Theory and Applications, 44(3), pp.209-252.
[8]Y.C. Hung, (2002) “Modeling and Analysis of Stochastic Networks with Shared Resources”, Ph.D.thesis, Department of Statistics, The University of Michigan.
[9]Y.C. Hung and G. Michailidis, (2007) “A Measurement Based Dynamic policy for Switched Processing Systems”, Proceedings of IEEE International Conference on Communications.
[10]Lawrence, E., Michailidis, G., Nair, V. N. and Xi, B. W. (2006), ”Network Tomography: A Review and Recent Developments”, in Frontiers in Statistics, Fan and Koul (eds.), Imperial College Press, London, pp. 345-366
[11]Tsang, Y., Coates, M. and Nowak, R. D. (2003), “Network Delay Tomography,” IEEE Transactions on Signal Processing, 8, pp 2125-2135
[12]Vardi, Y., (1996) “Network tomography: Estimation source-destination traffic intensities from source data,” JASA, 91, 365-377
[13]Duffield, N.G. and Lo Presti, F. (2004), Network Tomography from Measured End-to-End Delay Covariance, IEEE/ACM Transactions on Networking, 13, 325-336
[14]Lawrence, E., Michailidis, G., Nair, V. N. (2007) ”Statistical Inverse Problems in Active Network Tomography,” in Statistical Inverse Problems, R. Liu., W. Straderman, C. H. Zhang (editors), IMS Lecture Note Series.
[15]Castro, R., Coates, M, Liang, G., Nowak, R. and Yu, B. (2004), “Network Tomogra-
phy: Recent Developments,” Statistical Science, 19, 499-517