Pramod Viswanath

Position
Forrest G. Hamrick Professor in Engineering
Office Phone
Assistant
Office
B308 Engineering Quadrangle
Education
  • Ph.D. in Electrical Engineering and Computer Science, University of California-Berkeley, 2000
Advisee(s):
Bio/Description

Forrest G. Hamrick Professor in Engineering
Associated Faculty in Computer Science

I have two main research interests currently:

  1. First principle understanding and design of Blockchains.
  2. Inventing communication algorithms via Deep Learning

 

Full Stack Design of Blockchain Technologies

  • Teaching: Principles of Blockchains. full stack designs and lab implementations (University of Illinois-Urbana Champaign)
  • Research whitepapers:
    Distributed Payment Systems: Principles and Design. [PDF]
    Trifecta: The Blockchain Trilemma Solved. [PDF] [Short Video: Trilemma]
  • Prism: a consensus algorithm that meets information theoretic limits on throughput, latency, reliability -- a roughly 3 order improvement over Bitcoin, while maintaining same security guarantees. [Paper] [Short Video]
  • Coded Merkle Tree: a coded cryptographic hash aggregator that is key to scaling storage and communication efficiencies of blockchains. [Paper]
  • Spider: an off-chain networking stack bringing best practices of data networking to blockchains. [Paper]
  • Dandelion: redesigning Bitcoin networking for anonymity. Network-level anonymity properties in Bitcoin. [Paper] [Short Video]
  • Incentives: equitable designs for proof of stake blockchains. [Paper]

 

Inventing Communication Algorithms via Deep Learning

Emulating Viterbi and BCJR decoding via deep learning and harnessing the resultant neural networks to build robust and adaptive decoders for convolutional and Turbo codes for non-AWGN (bursty/fading) channels. [Paper]

The first family of codes in the presence of noisy feedback are designed via deep learning. While the problem is classical (initiated by Shannon) and feedback is known to improve practical performance, good codes are unknown and linear codes are known to be highly sub-optimal. Our neural network codes significantly outperform state of the art by several orders of magnitude in reliability. [Paper]

 

Past: Wireless Communication

I have worked actively in wireless communications for well over a decade, with research involving building models to designing algorithms to system solutions to entrepreneurial and industrial implementations. I am partial to the theoretical and conceptual aspects of the problems I work on. I was one of the early designers of Flash-OFDM, the first OFDM based cellular system (at Flatiron Technologies) and coauthored a popular book on wireless communication.

Selected Publications
  • see above for selected list

Google Scholar Profile

Honors and Awards:

  • Best Paper Award, Sigmetrics conference, 2015
  • Xerox Faculty Research Award, College of Engineering, UIUC, 2010
  • NSF CAREER Award, 2002
  • Eliahu Jury Award, UC Berkeley, EECS, 2000
  • Bernard Friedman Prize, UC Berkeley, Mathematics, 2000
Research Areas
Computing & Networking
Data & Information Science
Security & Privacy