scooter

Shashi Borade

Email: spb [at] mit [dot] edu

I am a graduate student in the EECS Department at Massachusetts Institute of Technology
and a member of the Laboratory for Information and Decision Systems (LIDS).

I work with Prof. Lizhong Zheng in the area of wireless communications and information theory.

Before coming to MIT in 2002, I was an undergraduate in the EE Department at IIT Bombay.

Update! I defended on June 24, 2008. Here are the slides.

Here are some particular topics of my interest (with sample publications).

Unequal error protection

Unequal error protection: some fundamental limits, with Baris Nakiboglu and Lizhong Zheng, submitted to IEEE Trans. on Info. Theory. 2008. (Slides)

Classical information theory ignores the issue of meaning for the engineering problem of communication and assumes all information to be equally improtant. In many situations however, all information is not created equal. We investigate some fundamental limits on how much extra protection can be provided to such high-priority information while still communicating the low-priority information reliably.

Using geometrical ideas in information theory

Euclidean information theory, with Lizhong Zheng, Allerton Conference, Sept. 2007. (Invited Paper)

Many problems in information theory involve optimizing the Kullback-Leibler divergence between probability distributions. Differential geometry motivates a local approximation of the KL divergence in terms of Euclidean distance. This Euclidean approximation simplifies KL divergence optimizations into linear algebra problems. Under this simplification, we solve the open problem of broadcast with degraded message sets for very noisy channels.

Network information theory

Multilevel broadcast networks, with Lizhong Zheng and Mitchell Trott, IEEE ISIT, June 2007.

Using graphical models in probability, we simplify the general problem of broadcasting with degraded message sets. This additional structure provides new insights for the general problem and solves it for a new set of situations. A converse result is provided in terms of a `mirror-image' of the actual network. The classical result for the physically-degraded situation is a simple corollary of this mirror-image converse.

Channels with feedback

Writing on fading paper and causal transmitter CSI, with Lizhong Zheng, IEEE ISIT, July 2006. (full version summitted to IEEE Trans. on Info. Theory)

Do not fight with the channel randomness, exploit it! A scheme motivated from dirty-paper coding is used for achieving capacity of a fading channel with causal state information (CSI) at the transmitter. The capacity per unit cost of a general channel with causal transmitter CSI is also derived using similar scheme.

Large deviations (error exponents)

Geometry of error exponents, with Lizhong Zheng, Allerton Conference, Sept. 2006. (Invited Paper)

A self-contained geometric perspective to various error exponents in information theory. It uses an intuitive approach based on a Pythagoras-like theorem for KL divergences. This shows the hidden geometry behind error exponents in information theory and provides new insights into the nature of rare events.

Complete list of Publications (with brief descriptions)

Curriculum Vitae (doc pdf)