Faculty Research Profiles


Click on a name to jump to their profile: Venkat Anantharam, Michael Gastpar, Kannan Ramchandran, Anant Sahai, David Tse, and Martin Wainwright.
Venkat Anantharam
Venkat Anantharam

Interests:


Theory; Communications; Control, Robotics, & Biosystems; Networks
Michael Gastpar
Michael Gastpar

Research Summary:


1. Network Information Theory

In an attempt to understand the behavior of information in general networks, we develop fundamental bounds on the coding performance both in source and in channel networks. One particular emphasis of our work concerns the scaling behavior as the number of nodes in the network becomes large. Recent results concern relay networks and distributed remote source coding (sometimes called the "CEO problem").

  • M. Gastpar and M. Vetterli, "On the Capacity of Wireless Networks: The Relay Case," In Proc IEEE Infocom 2002, New York, June 2002.


  • G. Kramer, M. Gastpar, and P. Gupta, "Cooperative Strategies and Capacity Theorems for Relay Networks," IEEE Transactions on Information Theory, Vol.51 No.9, pp.3037-3063, September 2005.

2. Distributed Signal Processing

The traditional signal processing paradigm is to ship all the data to one central location, where all the processing is done. Paradigmatic cases are all forms of transform processing. In the current key challenges (such as sensor and ad-hoc networks), the cost of the shipping step is a major bottleneck, and therefore, the as much as possible of the signal processing must be performed at the individual nodes, before shipping. In recent work, we have developed a distributed version of the Karhunen-Loeve transform (KLT).

  • M. Gastpar, P. L. Dragotti, and M. Vetterli, "The Distributed, Partial and Conditional Karhunen-Loeve Transforms," In Proc 2003 IEEE Data Compression Conference, March 2003.

3. Theories of Sensory Information

We develop theories of sensory information as it occurs in theoretical neuroscience and in the sensor nets that are currently being developed by the engineering community. Our theories exploit information-theoretic as well as signal processing arguments in order to develop appropriate abstractions. At their core, the theories address the absence of a "separation theorem": the universal abstraction of information in terms of "bits" is inappropriate, and a replacement must be found.

  • M. Gastpar and M. Vetterli, "Power, Spatio-Temporal Bandwidth, and Distortion in Large Sensor Networks," IEEE Journal on Selected Areas in Communications, Vol.23, No.4, pp.745-754, April 2005.


  • M. Gastpar and B. Rimoldi, "Cercal Sensory System: A Precise Sense In Which The Tuning Curves Are Optimized," Presented at the Conference on Computational and Systems Neuroscience (CoSyNe), Salt Lake City, UT, March 17-20, 2005.
Kannan Ramchandran
Kannan Ramchandran

Research Summary:


Kannan Ramchandran's research interests are at the broad intersection of signal processing, communications, information theory and networking. His current focus is on distributed and randomized signal processing and coding for wireless sensor networks, distributed architectures and coding algorithms for video-over-wireless and peer-to-peer networks, multi-terminal information theory and practical code constructions, multimedia security and information hiding, and multiscale image processing and wavelets.

Representative Publications:


  • R. Puri, A. Majumdar, P. Ishwar and K. Ramchandran, "Distributed Video Coding in Broadband Wireless Sensor Networks," To appear, IEEE Signal Processing Magazine, 2006.


  • D. Schonberg, S. C. Draper, and K. Ramchandran, "On Blind Compression of Encrypted Correlated Data Approaching the Source Entropy Rate," 43rd Annual Allerton Confrence, Monticello, IL, September 2005.


  • A. G. Dimakis, V. Prabhakaran and K. Ramchandran, "Ubiquitous Access to Distributed Data in Large-Scale Sensor Networks through Decentralized Erasure Codes", Symposium on Information Processing in Sensor Networks (IPSN), April 2005.


  • P. Ishwar, R. Puri and K. Ramchandran, "On rate-constrained distributed estimation in unreliable sensor networks", IEEE Journal on Selected Areas in Communications, Special issue on self-organized distributed collaborative sensor networks, April, 2005.


  • D. Petrovic, K. Ramchandran, J. Rabaey, "Overcoming Unreliable Radios in Sensor Networks with Network Coding," IEEE Workshop on Network Coding, Theory and Applications (NetCod), March 2005.


  • S. Sandeep Pradhan, Rohit Puri and Kannan Ramchandran, "n-channel symmetric multiple descriptions - Part I: (n,k) source-channel erasure codes," IEEE Transactions on Information Theory, Vol.50, No.1, January 2004.


  • P. Ishwar, V. M. Prabhakaran, and K. Ramchandran, "Toward a Theory for Video Coding Using Distributed Compression Principles," Proc. IEEE Int. Conf. Image Processing, Barcelona, Spain, September 2003.


  • P. Ishwar, A. Kumar, and K. Ramchandran, "Distributed Sampling for Dense Senor Networks: a ``bit-conservation principle''," Symposium on Information Processing in Sensor Networks, Proceedings of the Second International Workshop, Palo Alto, CA, April 2003.


  • S.S. Pradhan and K. Ramchandran, "Distributed Source Coding Using Syndromes (DISCUS)," IEEE Transactions on Information Theory, Vol.49, No.3, March 2003.


  • R. Puri and K. Ramchandran, "PRISM: A New Robust Video Coding Architecture Based on Distributed Compression Principles," Proc. Allerton Conf., Oct. 2002.
Anant Sahai
Anant Sahai

Research Summary:


1. Fundamental Limits for Cognitive Radios

We are exploring the basic constraints on the operation of opportunistic radios that take advantage of the gaps in spectrum utilization that currently exist, while preserving the operation of legacy systems. The goal is to provide the appropriate technical perspective for a roadmap for spectrum regulation in the medium to long term. Our initial results indicate that there might be substantial advantages to building out non-frequency-specific infrastructure (whether ad-hoc or otherwise). This suggests that the conventional wisdom behind exclusive property-rights in spectrum could be missing half the story since such a right to exclude would discourage, rather than encourage, investments in the potentially most useful infrastructure. So far, our technical focus has been on quantifying the limits on the sensing side. While in standard data communication, it is possible to engineer systems that typically operate at medium-to-high SNRs, the opportunistic case is fundamentally limited by low SNR behavior since that is the typical case when a band is free for opportunistic use! Opportunistic systems are thus more sensitive to uncertainties and unmodeled characteristics of the environment, and we have begin to quantify how cooperative diversity might be required to overcome some of these sensitivies. While many multiuser data communication systems are interference limited and require power control and MAC protocols to maintain data throughput, in the opportunistic case, power control and MAC protocols must also be used to preserve the sensitivity required to guarantee non-interference with potential legacy systems.

2. Delay Universality in Channel and Source Coding

Rather than fixing the message set and letting the block-length vary with the source, we explore fixing the input rate and letting the target delay vary with the application. This sort of delay universality turns out to be required in the relevant separation theorems for certain unstable processes. So far, we have explored these ideas in distributed lossless source coding, point-to-point channel coding, as well as multiterminal channel coding problems like MAC and degraded broadcast channels. The ultimate goal of the multiterminal work is to understand what the right architecture is for delay-sensitive and rate-sensitive applications to share a common communication medium. Preliminary results are already questioning the conventional wisdom holding that delay-sensitive applications need to operate within a clean band without external interference.

3. Feedback, Reliability, and Control

We are exploring the use of feedback to improve the reliability of communication systems. So far, we have developed a new upper bound (the "focusing bound") that quantifies the substantial improvements in the fixed-delay reliability function with feedback, and showed that this bound is asymptotically achievable for certain classes of channels given noiseless feedback by using appropriate flow control. This improved reliability with feedback is particularly surprising given that fixed-block reliability functions can achieve no such gain. The fixed-delay reliability functions with feedback are particularly important since delay-universal (anytime) communication problems are intimately connected to remote stabilization problems in which a control system must operate over a noisy communication link. This connection is just one example of our larger research program in which we attempt to build an appropriate heirarchy of communication problems partially ordered by the channel resources they require. In addition to the fixed-delay reliability functions, we are also interested in both fixed and variable block schemes. For fixed blocks, we have shown that generic Gaussian multiuser schemes can achieve arbitrarily high reliability improvements with feedback, as long as they operate under an average power constraint. For variable blocks, we have been exploring the limits on the feedback required to get reliability gains. So far, we have considered average-rate-limited noiseless feedback as well as high-quality, but noisy, feedback.
David Tse
David Tse

Research Summary:


The current research in my group focuses on understanding fundamental limits in wireless communication systems using tools from information theory and other areas. Specific topics include MIMO communication, cooperative radios, spectrum sharing, adhoc and sensor networks.
Martin Wainwright
Martin Wainwright

Research Summary:


Prof. Wainwright's research group focuses on problems in statistical signal processing, coding and communication theory, and large-scale multivariate statistical models. On-going research projects include:

  • analysis of decentralized protocols for statistical inference and signal processing in communication-constrained environments, with applications to sensor networks


  • development and analysis of provably effective methods for error-control coding and lossy data compression


  • message-passing algorithms and variational methods for statistical learning and estimation in large-scale graphical models




Representative Publications:


  • M. J. Wainwright and M. I. Jordan, "A Variational Principle for Graphical Models" in New Directions in Statistical Signal Processing, Edited by S. Haykin et al. MIT Press, Cambridge, 2005.


  • X. Nguyen and M. J. Wainwright and M. I. Jordan, "Decentralized Detection using Kernel Methods," IEEE Transactions on Signal Processing, Vol.53, No.11, pp.4053 - 4066, November 2005.


  • J. Feldman and M. J. Wainwright and D. R. Karger, "Using Linear Programming to Decode Binary Linear Codes," IEEE Transactions on Information Theory, Vol.51, No.3, pp.954-972, March 2005.


  • M. J. Wainwright and T. S. Jaakkola and A. S. Willsky, "A New Class of Upper Bounds on the Log Partition Function," IEEE Transactions on Information Theory, Vol.51, No.7, pp.2313-2335, July 2005.