PhD Defense: "Compression and Dispersive Information Routing for Networks"

Kumar Viswanatha

December 6th (Thursday), 11:00am
Harold Frank Hall (HFH), Room 1132 (Computer Science Conference Room)

This talk focuses on problems related to compression, routing and communication of correlated sources across a network. One main line of research is related to minimum cost joint compression and routing of correlated sources across a network. A new routing paradigm, called “dispersive information routing” (DIR) is introduced, wherein the intermediate nodes are allowed to split a packet and forward different subsets of the received bits of each of the forward paths. Information theoretic bounds on the minimum cost achievable using DIR are derived and it is demonstrated using simple examples that DIR asymptotically outperforms conventional routing techniques, with the gains growing unboundedly with the network size, for certain networks. Building on the encoding principles, a new encoding scheme is proposed for the L-channel multiple descriptions problem, involving a novel strategy called “combinatorial message sharing”, where a unique common codeword is sent within each subset of the descriptions. For a general class of source-distortion pairs (which includes a Gaussian source under mean squared error), the resulting achievable rate-distortion region strictly improves upon the most well known region for the problem. The last section of the talk addresses practical design challenges in distributed quantization for large scale sensor networks operating in severe channel conditions. A unified framework for the design of error/erasure resilient and complexity constrained distributed quantizers is proposed, where the decoder complexity is explicitly controlled during the design. Simulation results indicate substantial gains over other state-of-the-art techniques and provide strong evidence that the approach opens the door to practical deployment of distributed coding in real world sensor networks.

About Kumar Viswanatha:

photo of kumar viswanatha Kumar B. Viswanatha received his B.Tech. in electrical engineering in 2008 from the Indian Institute of Technology – Madras (IIT – Madras), Chennai, India and his M.S. in electrical and computer engineering in 2009 from University of California at Santa Barbara (UCSB), USA. He is currently pursuing his Ph.D. in the same department at UCSB. He interned as a quantitative associate in the equity volatility desk at Goldman Sachs Co., New York in 2011 and as a systems engineer in the modem team at Qualcomm Inc., Santa Clara in 2012. His research interests include information theory, estimation theory, multimedia compression and networking.

Hosted by: Professor Kenneth Rose