Abstract

Abstract: The problem of cooperative allocation among multiple secondary users to avoid collisions in a cognitive radio network is considered. The channel availability statistics are initially unknown to all the secondary users and are learnt via sensing samples. Distributed learning and allocation schemes which minimize the total regret of the users when compared with the ideal scenario with known availability statistics and centralized allocation are proposed. The first scheme assumes minimal prior information in terms of pre-allocated ranks for allocation while the second scheme is fully distributed and assumes no prior information at the users. The two schemes have sum regret which is provably logarithmic in the total number of access slots. A lower bound is derived for any learning scheme which is asymptotically logarithmic in the number of slots. Hence, our schemes achieve asymptotic order optimality in terms of regret in learning and allocation.

Biography: I hail from Mysore, India. I received her B.Tech in Electrical Engineering from the Indian Institute of Technology Madras in 2004 and my M.S. and Ph.D in Electrical and Computer Engineering from Cornell University in 2009. At Cornell, I worked at the adaptive communications and signal processing (ACSP) group with Prof. Lang Tong as advisor. I have spent summers at the networking technologies group, IBM Watson Research, Hawthorne, NY. I am currently a post-doctoral researcher at the Stochastic Systems Group working with Prof. Alan Willsky. I will be joining EECS Dept., University of California Irvine as an assistant professor starting July 2010.

My research interests are in the area of statistical signal processing, networking and information theory. I have worked on inference and learning of graphical models, scalable algorithms and asymptotic analysis. I am the recipient of the IEEE Signal Processing Society (SPS) 2008 Young Author award, 2007-08 IBM Fran Allen PhD fellowship, 2006-07 Google Anita Borg award finalist and the student paper award at the 2006 Intl. Conf. on Acoustic, Speech and Signal Processing (ICASSP).