Networks’ Challenge: Where Game Theory Meets Network Optimization
Asu Ozdaglar, Massachusetts Institute of Technology
The sharp drop in the cost of broadband access, the emergence of new wireless and wireline platforms and a host of new applications have transformed communication networks, both expanding their scale and making their presence in our lives evermore ubiquitous. Communication networks are not only much larger scale than before, but they are also organized differently, having emerged from the interconnection of a multitude of autonomous, independently operated administrative domains and are thus more complex and far more decentralized than those of just two decades ago. The analysis of these networks not only necessitates tools to model their large-scale and decentralized operation, but also needs to recognize the importance of the “human element” in their functioning. The mathematical and conceptual issues raised by modern communication networks are shared by several others, including transportation, energy, and social networks.
In this talk, I will give an overview of a newly emerging interdisciplinary research area for analysis and control of large-scale decentralized networks that combines elements from distributed convex optimization with game-theoretic (multi-agent) modeling. In the first part of the talk, I will focus on fundamental challenges in communication networks and present recent research on developing game-theoretic models and distributed optimization methods for resource allocation among heterogeneous agents in networks. In the second part of the talk, I will discuss issues of learning, information exchange and decision making by individuals holding decentralized information who are connected through a social network.