Âé¶¹´«Ã½AV

Report on the 2016 JTG / Âé¶¹´«Ã½AV IT Society Summer School
This is a report the 2016 Âé¶¹´«Ã½AV IT Society Summer School held at the Indian Institute of Science, Bangalore, India.
Sep 24, 2016
Group Photo_Thursday

The 2016 Joint Telematics Group / Âé¶¹´«Ã½AV Information Theory Society Summer School on Signal Processing, Communications and Networks was held at the Indian Institute of Science (IISc), Bangalore, during June 27 – July 01, 2016.

The Summer School series started in 2009 as an initiative of the Joint Telematics Group (JTG). It has been covering contemporary research topics in signal processing, communications, information theory, and networks and has been mainly aimed at students and young researchers from all over India. While the Schools have been held annually since 2009, Âé¶¹´«Ã½AV Information Theory Society’s involvement as a financial and technical co-sponsor began in 2014.

This year, we made a departure from previous summer school formats. In the past, the school comprised of two short courses, each given by a leading expert on some topic within the broad realm of communications, signal processing, information theory and networks. It would span four days. This year, we had three courses, each of eight hours of lecture. The duration of the entire school was five days.

The three short courses comprising the 2016 Summer School were taught by B. V. Rao, Adjunct Professor, Chennai Mathematical Institute, Chennai, India, Upamanyu Madhow, Professor, University of California, Santa Barbara, USA, and Erdal Arikan, Professor, Bilkent University, Ankara, Turkey. B. V. Rao’s lectures were on concentration inequalities, Upamanyu Madhow lectured on millimeter wave communication networks, and Erdal Arikan spoke about Polar Coding.

B. V. Rao started with an overview of fundamental inequalities in probability including those by Chebyshev, Cramer, Chernoff, Hoeffding, Azuma and McDiarmid. He discussed the connections with the Johnson-Lindenstrauss lemma, and went on to discuss the Effron-Stein lemma, and its applications in graph theory, the VC theory, etc. Also discussed were the Curie-Weiss lemma, log-Sobolev inequalities, Talagrand’s inequalities and their application to problems like the stochastic traveling salesman problem. The lectures concluded with a discussion of isoperimetric inequalities and their connection to Talagrand’s inequalities.

 Upamanyu Madhow began with an introduction to basic link-budget calculations involved in determining the feasibility of millimeter-wave communications and showed that the research domain is interesting with challenging questions and the mmWave technology a realizable one. He presented recent theory and algorithms developed for large antenna arrays including compressive estimation and super-resolution (in particular, the newtonized orthogonal matching pursuit algorithm). A highlight of the discussion was the review of the Ziv-Zakai bound and related estimation-theoretic fundamental limits. He presented the key ideas involved in developing networks when one can have highly directional links, both for mesh networks and picocells. Then, he discussed important signal processing issues that need to be tackled for high bandwidth communication, including the challenges in using 1-bit ADCs (or those with a small number of bits). He also presented results from data collected over a 1km^2 area in Manhattan, which experimentally demonstrated that mmWave can potentially offer 1000x the data throughput when compared with LTE. The lectures concluded with a presentation of recent advances in short-range mmWave radar applications and a discussion on various open issues in the area.

Erdal Arikan started with a gentle introduction to the basic ideas in information theory such as the entropy, mutual information, discrete memoryless channels and the channel coding theorem.  Then he discussed the fundamental idea behind channel polarization, namely channel combining and splitting, the conservation of capacity by such an operation. He then discussed about low complexity polarization, showed its recursive extension and presented the main polarization theorem from his 2007 paper. He discussed the encoding and decoding complexity in detail and presented several examples. Then, he discussed the performance of polar coding and compared it with the state-of-the-art codes. He presented different options for decoding including maximum likelihood, successive cancelation, belief propagation, list decoding and sphere decoding, and discussed their performance-complexity tradeoff. He also discussed practical aspects: implementation performance measured in terms of chip area, throughp