Date: 10:00am, March 7, 2016
Location: Costa Commons, CEPSR 750
Speaker: Dr. Yucel Altug, Princeton University
Abstract: The availability of unprecedented amount of data has the potential to improve our lives. Efficient processing and transmission of massive data are significant obstacles to achieve this goal. Feedback information theory, thanks to its close relation to optimal learning problems and successful utilization in practical communication systems, is a fertile area of research to provide required notions and tools to address these challenges. However, understanding about feedback information theory is mostly incomplete with various longstanding open problems. In this talk, I will explain a new perspective to tackle the open problems regarding the effect of feedback on fundamental limits of communication systems. Specifically, I will show that feedback can improve the speed of convergence to the capacity for a certain class of discrete memoryless systems. This unexpected result is the very first demonstration of an improvement of an asymptotic fundamental limit of a memoryless communication system due to feedback. More importantly, it leads to a viable research direction: delineation of channels for which feedback improves a fundamental limit. As a concrete example, I will show the complete characterization of the effect of feedback on the asymptotic fundamental limits for the classical channel coding per unit cost problem. Finally, besides initiating a research program to understand the effect of feedback on memoryless communication systems, the insights from these results have also interesting implications for optimal learning problems, which I will briefly highlight.
Biography: Yucel Altug is a postdoctoral research associate in the Electrical Engineering Department at Princeton University hosted by Vince Poor and Sergio Verdu. He received the B.S. and M.S. degrees in Electrical and Electronics Engineering from Bogazici University, Turkey, in 2006 and 2008, respectively and the Ph.D. degree in Electrical and Computer Engineering from Cornell University, in 2013. He has received the ECE Director's Ph.D. Thesis Research Award from Cornell University School of Electrical and Computer Engineering. His research interests include information theory, feedback communications and optimal learning.
Hosted by Professor John Wright