IMS
Session Slot: 10:30-12:20 Thursday
Estimated Audience Size: 125-175
AudioVisual Request: Two Overheads
Session Title: Probability and Statistics in Telecommunications
Theme Session: No
Applied Session: No
Session Organizer: Koshevnik, Yuly MCI
Address: 2400 North Glenville Road, Richardson, TX 75082
Phone: (972) 918-6823
Fax: (972) 918-6257
Email: Yuly.Koshevnik@mci.com
Session Timing: 110 minutes total (Sorry about format):
Opening Remarks by Chair - 5 minutes First Speaker - 20 minutes Second Speaker - 20 minutes Third Speaker - 20 minutes Fourth Speaker - 20 minutes Discussant - 15 minutes (or none) Floor Discusion - 5 minutes
Session Chair: Koshevnik, Yuly MCI
Address: 2400 North Glenville Road, Richardson, TX 75082
Phone: (972) 918-6823
Fax: (972) 918-6257
Email: Yuly.Koshevnik@mci.com
1. Moving Images: Semi- and Nonparametric Approach
Korostelev, Alexander, Wayne State University
Address: Department of Mathematics, F.A.B. 1247, Wayne State University, Detroit, MI 48202
Phone: (313) 577-3188
Fax: (313) 577-7596
Email: apk@math.wayne.edu
Abstract: On the basis of some examples, we discuss the peculiarities of the asymptotical large-sample analysis in nonparametric models of the moving images. As the asymptotical rates of convergence show, these problems cannot be treated as ``snap-shot'' images in the ``time-space'' domains.In the semiparametric image models, we want to estimate a finite dimensional parameter from noisy data in presence of an infinite dimensional ``nuisance parameter'' which describes the image shape. For example, if the evolution of the image is governed by some dynamical system, how accurately can the parameters of this system be restored? Unlike the classical semiparametric regression and density problems in which the root-n rates are typical under regularity conditions, in the image analysis the accuracy in estimating the finite dimensional parameter relates to the smooth functionals estimation. The smooth functionals of image such as the area, the center of gravity, etc., are known to have the intermediate rates of convergence comparing to those in parametric and nonparametric estimation. That's why we find the semiparametric image problems different from their parametric and nonparametric counterparts.
2. Optimization of Polling Systems with Fast Service
Kreinin, Alexander, Algorithmics, Inc.
Address: Algorithmics Inc. 822 Richmond Street West, Toronto, ON, Canada, M6J 1C9
Phone: (416) 703-0898
Fax: (416) 703-0767
Email: alex@algorithmics.com
Abstract: Polling systems with non-zero switchover time, intensive arrival flows, fast service and random discipline of service are studied in the talk. Under some assumptions, the process of queue lengths convergence to the trajectories of a dynamical system in the N-dimensional Euclidean space. We study the asymptotic behavior of the process, describe the properties of the limit dynamical system and obtain formulae for mean queue lengths expressed through the first moments of the input flow distributions, the distributions of service times and switchover times. We also derive the optimal random routing policy to minimize the average number of customers in the polling system.
3. Monitoring of High Intensity Data Streams
Yashchin, Emmanuel, IBM Research Center
Address: Thomas J. Watson IBM Research Center, P.O. Box 218, Yorktown Heights, NY 10598
Phone: (914) 945-1828
Fax: (914) 945-3434
Email: yashchi@watson.ibm.com
Abstract: This talk will discuss several problems related to analysis of data streams subject to abrupt changes in time (shifts, onset of drifts, etc.). Such data is frequently observed in a wide range of areas, such as quality control, communications and finance. It is typically described in terms of regimes and parameters that can be estimated, controlled or monitored. The talk will focus on the problems of detection of unfavorable changes, estimation of the current level of parameters (filtering) and retrospective data analysis. It will present methodology based on the change-point theory for addressing these problems and give several examples of its application.
4. Interpretation Aids for Data-Rich Environments
Hardy, William C., MCI Telecommunications Corp.
Address: 2400 North Glenville Road, Richardson, TX 75082
Phone: (972) 918-5925
Fax: (972) 918-6257
Email: Chris.Hardy@mci.com
Abstract: A data-rich environment is understood here to be one in which there is so much readily available data that it is possible to create samples with virtually no sampling variance. Such an environment is typified by the long-distance telecommunications industry, in which sample sizes of hundreds of thousands of links, millions of calls, billions of equipment operating hours, trillions of bits transmitted, etc. are the norm rather than the exception. This paper describes the shift in concerns from statistical to operational significance that occurs in such an environment, and illustrates analytical techniques that have been found to be useful in addressing those operational concerns with examples drawn from the author's experience as a telecommunications operations analyst.
Discussant: Koshevnik, Yuly MCI
Address: 2400 North Glenville Road, Richardson, TX 75082
Phone: (972) 918-6823
Fax: (972) 918-6257
Email: Yuly.Koshevnik@mci.com
List of speakers who are nonmembers: None