ECE Colloquium with Vahid Tarokh

Information in Non-Linear Regime With Professor Vahid Tarokh, Division of Engineering and Applied Sciences, Harvard University Refreshments will be served outside Room 339 at 1:45 p.m. Faculty Host: Bobak Nazer Abstract: Communications devices (e.g. power amplifiers and mixers) are inherently nonlinear, and most designers try to make/use devices that manifest almost linear behaviors in certain regions of interest. The operations of communications devices are then limited to the linear regions. There have been some studies of communications and signal processing in the nonlinear regime, most notably for compensation of amplifier non-linearities. In spite of this, information theory of nonlinear communications is in a primitive stage. A similar assessment can be made of communications and signal processing (for communications) in this regime. We will begin with our recent results on the capacity of non-linear channels. The results indicate that typical compressive non-linearities limit the system capacity at high signal to noise ratios. We recover Shannon's famous capacity formula when we approach the linear regime. We will then discuss a number of remaining open problems. As an example of our efforts in signal processing, we will consider the identification of nonlinear channels represented by an unknown but small number of dominant non-linear modes (e.g. Volterra series terms), and talk about the development of a number of recursive online algorithms for adaptive identification of such non-linear channels. These algorithms provide significant improvement over the conventional algorithms even when restricted to linear regime, both in terms of mean squared error (MSE) and computational complexity. Gains in the order of 19 dB can be achieved at much lower complexities with an online implementation in some typical cases. About the Speaker: Vahid Tarokh received the Ph.D. degree in Electrical Engineering in 1995. He worked at AT&T Labs-Research and AT&T Wireless Services until 2000 where he was (in chronological order) Senior Member of Technical Staff, Principal Member of Technical Staff and Head of Department of Wireless Communications and Signal Processing. In 2000, he joined the Electrical Engineering Department at MIT, where he taught for two years as an Associate Professor. He then joined Harvard’s faculty and was appointed a Gordon MacKay Professor of Electrical Engineering in 2002. He was named Perkins Professor and Vinton Hayes Senior Research Fellow of Electrical Engineering in 2005. His most recent awards are a 2011 Guggenheim Fellowship, 2012 IEEE TCCN (Technical Committee on Cognitive Networks) Publication Award, and a 2013 IEEE Eric E. Sumner Award.