ECE Seminar with Georgios Piliouras

Competing Algorithms With Georgios Piliouras Research Scientist Georgia Institute of Technology Faculty Host: Yannis Paschalidis Refreshments will be served outside Room 339 at 1:15 p.m. Computation is becoming increasingly ubiquitous and interconnected. Algorithms do not exist in isolation any more but interact with each other. These computational ecosystems raise significant analytical challenges as we need to understand and quantify tradeoffs between efficiency, stability, and robustness to environmental uncertainty. In this talk, I will show how to address such issues by combining notions from game theory and nonlinear dynamical systems inside an algorithmic/optimization framework. By analyzing stability of equilibria and quantifying the size of their basin of attraction we can prove strong system guarantees that in some cases outperform even best case equilibria. Furthermore, by introducing elements from robust/stochastic optimization we can extend such tight, approximately optimal performance guarantees even under systemic uncertainty. Finally, I will discuss a novel framework for analyzing persistent nonequilibrium patterns in distributed systems. About the Speaker: Dr. Piliouras is a research scientist at the Georgia Institute of Technology (GaTech). He received his PhD in Computer Science from Cornell University in 2010. Before that he received a diploma in Electrical Engineering and Computer Science from the National Technical University of Athens (2004), and a MSc in Mathematical Logic and Algorithms from the University of Athens (2005). He has held visiting positions at the University of Oxford, UK (Dept. of Economics) and CWI (National Research Institute for Mathematics and Computer Science, Netherlands) and has been an invited lecturer in international summer schools on algorithms and game theory. He is the recipient of Olin Fellowship (Cornell). His research focus is algorithmic game theory, which lies at the intersection of algorithm design and game theory and whose objective is to design algorithms suitable for strategic environments. In his work, he has applied these principles to a number of different areas including micro-economic theory, distributed routing, control of prototype neuromorphic microchips, robotic systems as well as information theory.