Kartik Venkat to be awarded 2016 Thomas M Cover Dissertation Award
Kartik Venkat, Stanford University, PhD. is announced as the recipient of the Thomas M. Cover Dissertation Award for 2016, from the IEEE Information Theory Society, for his dissertation obtained under the direction of Professor Tsachy Weissman in the Information Systems Lab of the Stanford Electrical Engineering Department.
Jun 10, 2016
KartikVenkat

Kartik Venkat, Stanford University, PhD. is announced as the recipient of the Thomas M Cover Dissertation Award for 2016, from the IEEE Information Theory Society, for his dissertation obtained under the direction of Professor Tsachy Weissman in the Information Systems Lab of the Stanford Electrical Engineering Department. Kartik Venkat's Dissertation defended December 2015 is titled:

Relations between information and estimation: A unified view

This thesis considerably extends and unifies a topic at the intersection of information theory, probability, statistical signal processing, communication and control, that has been given much attention, largely due to an influential paper by Guo, Shamai and Verdu' in the transactions of our society in 2005, with precursors in treatments of causal estimation in the 1970s and the entropic central limit theorem in the 1980s.  These precursor results relate information quantities such as entropy and mutual information to integrated mean squared errors of estimation.

Three of Kartik Venkat's dissertation results are here highlighted. One is Kartik's discovery of point-wise relationships between log-likelihood ratios and integrated squared error loss, as random quantities, with the aforementioned results arising via mathematical expectation. A second result is the extension of these identities to continuous-time stochastic processes via martingale identities and stochastic calculus, fitting in a probabilistic framework of appropriate generality called Le'vy channels.  Estimation and information characterization for gamma and negative binomial channels are worked out as examples, with determination of the right notions of loss (extending the squared error loss) and determination of the appropriate notion of signal-to-noise ratio (SNR) for these channels.  Kartik's thesis also explored the three natural estimation tasks of filtering, finite look-ahead prediction, and infinite look-ahead smoothing, showing that while the filtering and smoothing mean squared errors for the Gaussian channel have mutual information rate characterization, the finite look-ahead prediction error does not.

It is noteworthy that in addition to the Cover Award winning Dissertation, Kartik Venkat has eight published journal papers, six of which are in the IEEE Transactions on Information Theory, one in the IEEE Transactions on Signal Processing, and one in Genomics published by BioMed Central. Plus Kartik Venkat was previously recognized for the outstanding student paper at the 2012 International Symposium on Information Theory for part of the work that subsequently went into his dissertation.

Kartik Venkat will receive the Thomas M. Cover Dissertation Award at the International Symposium on Information Theory in Barcelona in July 2016.