 |
Center for Nonlinear Studies |
Networking or networks is a topic that has attracted intense research activity spanning various disciplines since the late 1980's, and it remains today an active area of research in diverse fields such as computer science, communications, control, biological sciences, and the social sciences. Among the different types of networks that exist today, communication and computer networks as well as social networks enjoy the forefront interest, driven by the challenging problems they lead to in several of these disciplines, and by the technological revolution created by the Internet. Underlying issues range from flow control, routing, congestion control, pricing of services and mechanism designs for allocation of divisible resources to evolution of preferences and consensus formations. A broad spectrum of tools from various disciplines are needed to address the challenges offered by the topical issues listed above, and quite often it is a judiciously selected blend of tools from across disciplines that has the most impact in any particular problem area. One such rich set of tools, of both conceptual and algorithmic nature, is provided by game theory, specifically in settings where competition and/or cooperation among different entities are commonplace, where there is also the need for strategic decisionmaking. Another discipline that provides additional complementary tools for networking, within market structures, is economics, with the close relationship (between networking and economics) leading to a fairly new discipline called network economics. Recent research activities in networks have seen the convergence of what used to be disparate fields---of networks, games, and economics. It is this relatively recent confluence that this talk will address, dwelling on a games/economics-centric framework for networking, with particular focus on concepts, methodology, modeling, and tools for implementation, within the context of communication and computer networks. It will discuss various issues that impact networking research and its applications, with two main topics being pricing of services and incentivizing through mechanism designs (or reverse engineering). Both these topics involve multi-agent decentralized decision-making in an uncertain environment with distributed computation. The talk will cover some specific existence, uniqueness and computational issues that arise in these contexts, as well as some asymptotics for large-population regimes.
Host: David Wolpert, CCS-3, 665-7914, Game Theory Seminar Series