Electrical Engineering

by Dr. Rahul Jain

IBM Watson Research 

When: Monday, March 17, 2008

Time: 11:00 a.m. – 12:00 p.m.

Location: A265 Bourns Hall 

Abstract:   

With the increasing scale and complexity of networks, decentralization mechanisms become very important. Examples are the Internet and sensor networks. Often, these involve private incentives since parts of the network are owned by different entities, or the nodes may behave independently such as in sensor networks. For such problems, we would like to develop decentralized algorithms that involve minimal coordination, and in fact, may also align the private (and sometimes strategic) incentives of such entities. Ideas of decentralization from pricing and game theory have proved very fruitful in many network problems such as congestion control and routing.

I will begin the talk by giving a quick overview of the work on network pricing and games. We will then consider a particular problem, namely allocation or exchange of network resources between network service providers and users. We will see how market mechanisms can be used for solving such problems. It can be shown that the combinatorial market mechanism that I propose has very nice
properties including a zero price of anarchy (or full efficiency), i.e., every Nash equilibrium yields the efficient allocation in the complete information case.  In the case of incomplete information, we show the mechanism is asymptotically Bayesian incentive-compatible.  We will see its application to large scale networks such as the sensor networks, and in particular, the Internet.

 About the Speaker:  

Rahul Jain received his B.Tech from the Indian Institute of Technology, Kanpur in 1997, an MS in ECE from Rice University in 1999, and an MA in Statistics in 2002 and a PhD in EECS in December 2004, both from the University of California. He has been a recipient of a Texas Instruments Graduate Fellowship, and an IBM Invention Achievement Award.