Faculty Candidate Seminar: Statistical Inference for Network Tomography

GUEST LECTURER
Gang Liang

AFFILIATION
University of California @ Berkeley

ABSTRACT
Abstract: We study network tomography questions, which are large scale inverse
inference problems and vital for many network traffic engineering tasks such as
dynamic routing optimization, Quality of Service (QoS) guarantee. We propose a
pseudo likelihood approach for estimating parameters of these problems based on
the principle of divide-and-conquer. The example of multicast link delay
estimation problem is used to motivate the concept. We then apply the pseudo
likelihood approach to the problem of estimating origin-destination matrix
through link traffic counts, which is one of the core problems in network
traffic engineering.

This is joint work with Bin Yu.

DATE & TIME
Monday, February 16, 2004 -- 10:00 AM

DURATION
1 hour

LOCATION
Executive Classroom

CONTACT PERSON
roshan@isye.gatech.edu