Bombay Information Theory Seminar 2020

The third Bombay Information Theory Seminar (BITS) will be held during January 6-9, 2020 in Mumbai. The first of this biennial series was held in 2016 to commemorate the birth centenary of Claude Shannon. BITS 2020 will feature three days of talks and a tutorial.

Speakers

(click on the photos for abstracts & videos)

Venkat Anantharam

UC Berkeley

Himanshu Asnani

TIFR

Adrish Banerjee

IIT Kanpur

Srikrishna Bhashyam

IIT Madras

Avhishek Chatterjee

IIT Madras

Tanvi Jain

ISI Delhi

Varun Jog

Wisconsin

Animesh Kumar

IIT Bombay

Chandra Murthy

IISc

Parimal Parag

IISc

Sundar Rajan

IISc

Lalitha Sankar

Arizona State

Pranab Sen

TIFR

Devavrat Shah

MIT

Nihar Shah

CMU

Vinod Sharma

IISc

Rajesh Sundaresan

IISc

Vincent Tan

NUS

Andrew Thangaraj

IIT Madras

Himanshu Tyagi

IISc

Rashmi K. Vinayak

CMU

Michèle Wigger

Telecom ParisTech

Schedule

(click on a talk for its abstract & video)

12:00

Refreshments
pizza, fruit, ice cream
13:30 - 14:20

Convertible codes: New class of codes for efficient conversion of coded data in distributed storage
Rashmi K. Vinayak
14:20 - 14:50

Learning to detect an odd Markov arm: the restless case
Rajesh Sundaresan
14:50

Tea
15:10 - 15:40

Inference of spatial fields with a location-unaware mobile sensor
Animesh Kumar
15:40 - 16:25

Extracting robust and accurate features via a robust information bottleneck
Varun Jog

9:30 - 10:15

Variations on the theme of synthetic control
Devavrat Shah
10:15 - 10:45

A method to find the volume of a sphere in the Lee metric, and its applications
Adrish Banerjee
10:45

Tea + Posters
11:30 - 12:15

MIMO free-space optical channels
Michèle Wigger
12:15 - 12:45

Downlink Pareto optimal beamforming with limited cooperation
Srikrishna Bhashyam
12:45 - 13:15

Achievable rates for a finite block length in a fading MAC
Vinod Sharma
13:15

Lunch
14:15 - 15:00

Community detection and matrix completion with two-sided graph side-information
Vincent Tan
15:00 - 15:30

Convergence of Chao unseen species estimator
Andrew Thangaraj
15:30

Tea + Posters
16:00 - 16:30

Coded caching with error correction
B. Sundar Rajan
16:30 - 17:00

Tracking an auto-regressive process with limited communication
Parimal Parag
17:00

Panel Discussion

Women in Research: Challenges in Space and Time
Organizer: Lalitha Vadlamani

9:30 - 10:15

Battling demons in peer review

Nihar Shah
10:15 - 10:45

Observability and controllability of linear dynamical systems under sparsity constraints
Chandra Murthy
10:45

Tea + Posters
11:30 - 12:15

Alpha-loss: A tunable class of loss functions for robust and fair learning
Lalitha Sankar
12:15 - 12:45

CCMI : Classifier based conditional mutual information estimator
Himanshu Asnani
12:45 - 13:15

Domain compression: A new primitive for information constrained statistics
Himanshu Tyagi
13:15

Lunch
14:15 - 15:00

One-shot Information Theory via the Poisson matching lemma
Venkat Anantharam
15:00 - 15:30

Scheduling with guaranteed quality of experience
Avhishek Chatterjee
15:30

Tea + Posters
16:00 - 16:45

Convexity properties of the sandwiched quasi-relative entropy
Tanvi Jain
16:45 - 17:15

Unions, intersections and a one shot quantum joint typicality lemma
Pranab Sen

9:30 - 13:00

Tutorial: Resilient and efficient distributed storage and computation via coding theoretic tools
Rashmi K. Vinayak

VENUE

TIFR

Jan 6

Homi Bhabha Auditorium

Getting to TIFR

IIT

Jan 7-9

LH 102, 1st Floor, Lecture Hall Complex (campus map)

Getting to IIT

ORGANIZATION

Vivek Borkar

Bikash Dey

Nikhil Karamchandani

Sibiraj Pillai

Vinod Prabhakaran

Jaikumar Radhakrishnan

Large-scale systems often operate in scenarios where they are prone to non-ideal conditions such as failures, unavailabilities, stragglers, load imbalance, etc. These issues adversely affect data durability and access latency in storage systems and query latency in computation systems. Redundancy is a common approach employed to impart resilience and mitigate these adverse effects. While replicating data and computation is a simple approach to provide resilience, it entails significant resource overheads. Coding theory based tools have emerged as good candidates to add redundancy in a resource efficient manner.

In the first part of the tutorial I will focus on distributed storage. Erasure coding based redundancy has already found widespread adoption in large-scale storage systems, and recent research from the coding theory community has had real-world impact. I will first present an overview of these works. Then I will discuss an exciting new opportunity in coding for distributed storage. This relates to a new approach for redundancy-configuration in storage systems which opens up a number of challenging and potentially impactful problems in coding theory.

In the second part of the tutorial I will focus on distributed computation on modern infrastructures. Coded computation has emerged as a resource-efficient way of adding resilience to distributed computation, wherein multiple units of data are encoded to create parity units and the function to be computed is applied to each of these units on distinct servers. If some of the function outputs are unavailable, a decoder can use the available ones to decode the unavailable ones. I will present recent research in this area, which are quite nascent with a number of exciting open problems.

Erasure codes are typically used in large-scale distributed storage systems to provide durability of data in the face of failures. In this setting, a set of k blocks to be stored is encoded using an [n, k] code to generate n blocks that are then stored on different storage nodes. A recent work by Kadekodi et al. (2019) shows that the failure rate of storage devices vary significantly over time, and that changing the rate of the code (via a change in the parameters n and k) in response to such variations provides significant reduction in storage space requirement. However, the resource overheads of realizing a change in the code rate on already encoded data in traditional codes is prohibitively high.

Motivated by this application, in this work, we first present a new framework to formalize the notion of code conversion---the process of converting data encoded with an [nI, kI] code into data encoded with an [nF, kF] code while maintaining desired decodability properties, such as the maximum-distance-separable (MDS) property. We then introduce convertible codes, a new class of code pairs that allow for code conversions in a resource-efficient manner. For an important parameter regime, along with the widely used linearity and MDS decodability constraints, we prove tight bounds on the number of nodes accessed during code conversion. We also present explicit low-field-size constructions of access-optimal MDS convertible codes for a broad range of parameters.

This is the 2020 Prof. R. Narasimhan Endowment Award Lecture.

Bombay Information Theory Seminar 2020

Speakers

(click on the photos for abstracts & videos)

Venkat Anantharam

Himanshu Asnani

Adrish Banerjee

Srikrishna Bhashyam

Avhishek Chatterjee

Tanvi Jain

Varun Jog

Animesh Kumar

Chandra Murthy

Parimal Parag

Sundar Rajan

Lalitha Sankar

Pranab Sen

Devavrat Shah

Nihar Shah

Vinod Sharma

Rajesh Sundaresan

Vincent Tan

Andrew Thangaraj

Himanshu Tyagi

Rashmi K. Vinayak

Michèle Wigger

Schedule

(click on a talk for its abstract & video)

Convertible codes: New class of codes for efficient conversion of coded data in distributed storage

Learning to detect an odd Markov arm: the restless case

Inference of spatial fields with a location-unaware mobile sensor

Extracting robust and accurate features via a robust information bottleneck

Variations on the theme of synthetic control

A method to find the volume of a sphere in the Lee metric, and its applications

MIMO free-space optical channels

Downlink Pareto optimal beamforming with limited cooperation

Achievable rates for a finite block length in a fading MAC

Community detection and matrix completion with two-sided graph side-information

Convergence of Chao unseen species estimator

Coded caching with error correction

Tracking an auto-regressive process with limited communication

Women in Research: Challenges in Space and Time

Battling demons in peer review

Observability and controllability of linear dynamical systems under sparsity constraints

Alpha-loss: A tunable class of loss functions for robust and fair learning

CCMI : Classifier based conditional mutual information estimator

Domain compression: A new primitive for information constrained statistics

One-shot Information Theory via the Poisson matching lemma

Scheduling with guaranteed quality of experience

Convexity properties of the sandwiched quasi-relative entropy

Unions, intersections and a one shot quantum joint typicality lemma

Tutorial: Resilient and efficient distributed storage and computation via coding theoretic tools

VENUE

TIFR

Jan 6

Homi Bhabha Auditorium

Getting to TIFR

IIT

Jan 7-9

LH 102, 1st Floor, Lecture Hall Complex (campus map)

Getting to IIT

ORGANIZATION

Vivek Borkar

Bikash Dey

Nikhil Karamchandani

Sibiraj Pillai

Vinod Prabhakaran

Jaikumar Radhakrishnan