[UCI-Calit2] Strategic Learning and Dynamics in Network and Computing Games

Anna Lynn Spitzer aspitzer at calit2.uci.edu
Thu Feb 14 08:49:38 PST 2008


Strategic Learning and Dynamics in Network and Computing Games
A Networked Systems Distinguished Speaker Event


With Mihaela van der Schaar, associate professor Department of
Electrical Engineering, UCLA

2-3 p.m.
Thursday, Feb. 14
Calit2 Room 3008

Networked devices and systems currently operate in a passive manner that
limits both the individual and the overall network/system performance.
Specifically, they try to maximize their immediate utility by myopically
optimizing their operation, without trying to forecast the impact of
their actions on the future utilities, or to proactively influence the
long-term network/system dynamics. 
In this talk, we propose to fundamentally change the passive way in
which devices currently interact, by enabling them to proactively
compete for the dynamic resources and improve their performances, based
on their information, learning capabilities and available actions.
Specifically, we introduce a new networking and computing paradigm,
where devices compete in various centralized and decentralized resource
markets regulated by users, network providers, or system managers,
aiming to maximize their own utilities or the social welfare. To model
and design various resource markets, we propose a general stochastic
game formulation. When operating in such resource markets, devices
become selfish, autonomous users, who strategically interact in order to
acquire the necessary resources for optimizing their performance, given
the experienced network, system, and application dynamics. The actions
selected by the devices for playing the stochastic game are the
cross-layer algorithms and parameters that can be adopted at the various
layers of the protocol stack or system layers. To be able to make
foresighted and proactive decisions, users will need to learn and model
directly or indirectly the other users' responses to their actions. We
study the outcome of various dynamic interactions among strategic users
possessing different knowledge levels and actions, and show that the
proposed paradigm can lead to multi-user systems that achieve new
measures of optimality, rationality and fairness. Exemplary systems to
which the proposed paradigm was already successfully applied include
multi-user competitions in existing wireless network infrastructures,
emerging cognitive radio networks, peer-to-peer networks and multi-task
processing systems. Our results show that smart users, which are able to
successfully learn, forecast, negotiate, interact and adapt, will not
only be able to derive higher utilities, but also help to achieve
unprecedented network or system efficiency improvements. 


Mihaela van der Schaar is currently an associate professor in the
Electrical Engineering Department at UCLA. She received the NSF Career
Award in 2004, the Best Paper Award from IEEE Transactions on Circuits
and Systems for Video Technology in 2005, the Okawa Foundation Award in
2006, the IBM Faculty Award in 2005 and 2007, and the Most Cited Paper
Award from EURASIP: Image Communications journal in 2006. She holds 30
granted U.S. patents and three ISO awards. Her research interests are in
multimedia communications, networking, architectures, systems,
compression and processing. For more information on her research, please
visit http://medianetlab.ee.ucla.edu.

For more information, contact Athina Markopoulou, athina at uci.edu. 


More information about the UCI-Calit2 mailing list