I am a postdoctoral research associate with the Center for Science of Information - an NSF funded science and technology center located at Purdue University.
I am mentored by Prof. P R Kumar of Texas A & M University and Prof. Wojciech Szpankowski of Purdue University. I am hosted at Purdue University.
Current Research Interests : Privacy, Data Science, Information Theory, Network communication.
Please contact me if you wish to speak in our CSoI seminar series. Please find the schedule here.
(26 Jan 2018) -- My accepted journal article on three-user broadcast channel is on IEEE Early Access. Preprint available here.
(12 Jan 2018) -- Submitted my privacy work to ISIT 2018.
(6 Nov 2017) -- Submitted my findings on Joint-source channel coding to IEEE Transactions on Information Theory. Preprint available in arxiv.
(1 Nov 2017) -- On the Discrete Geometry of Differential Privacy available here.
(25 June 2017) -- Presented my work on Communicating Correlated Sources over MAC at ISIT 2017, held in Aachen, Germany.
(27 June 2017) -- Presented my work on Communicating Correlated Sources over Interference Channel at ISIT 2017, held in Aachen, Germany.
(4/1/2017) -- Communicating Correlated Source Over MAC accepted at ISIT 2017.
(4/1/2017) -- Communicating Correlated Source Over Interference Channels accepted at ISIT 2017.
Research and Teaching Experience
Here is a brief summary of research and teaching activities during my doctoral and post-doctoral years in reverse chronological order.
Postdoctoral research : In May 2015, I was awarded the NSF postdoctoral fellowship by CSoI to conduct research in the broad areas of data science and Information Theory. I have been conducting research in this role since August 2015 and I have been interacting with multiple faculty members of CSoI. In particular, I have interacted very closely with Prof. P R Kumar and Prof. Wojciech Szpankowski.
The first thread of my current research is in the area of privacy-preserving data analysis. As part of this initiative, I have proposed an architecture for extracting statistical information from databases subject to privacy constraints. The architecture is aimed to overcome the problem of accumalated information due to repeated querying of databases. I have characterized a fundamental utility-privacy tradeoff that governs the performance of this architecture. Characterizing this utility-privacy tradeoff boils down to solving an important open problem in differential privacy. My study of this problem led me to unravel rich combinatorial structure of the underlying optimization problem. Leveraging sophisticated tools from discrete geometry, combinatorics and analytic methods, I provided a complete solution to this tradeoff via a simple closed form computable expression. In particular, I have proved that the above mentioned utility-privacy tradeoff can be obtained as simple functional of the Ehrhart series of a suitably defined convex polytope. The latter is a fundamental construct in Ehrhart theory and my findings establishes a connection between differential privacy and Ehrhart Theory. My findings are described in my preprint, which is in preparation for submission to IEEE Transactions on Information Theory.
The second thread of my research is aimed at designing efficient strategies for enabling communication of distributed correlated information. Systems and technologies that we are currently building, such as Internet of Things (IoT), fleet of autonomous vehicles or the smart grid rely on gathering and communicating correlated information originating at distributed terminals. How do we efficiently utilize correlation among these distributed information sources to communicate efficiently over noisy channels? This question has remained open for several decades now with the current known best strategies dating back to the 1980s. As part of my postdoctoral research, I have developed a new strategy to effect efficient communication of correlated sources over general multiple access and interference channels. Characterizing its information theoretic performance, I have proved that the developed strategy strictly outperforms all previous known startegies. This yields a new inner bound to the problem of joit source-channel coding over MAC and interference channels. These findings have been published in my papers at ISIT 2016 and ISIT 2017 and ITA 2017. I have compiled these findings and submitted them to Transactions on Information Theory. Preprints are available here and here. It maybe noted that these are single-author works.
From Nov 2014 to Aug 2015, I worked for Ericsson. Inc. at San Jose, CA as a Research Engineer in their Radio Access Research team.
Doctoral research and activities : I received my doctorate from the Department of Electrical Engineering and Computer Science (EECS), University of Michigan, Ann Arbor (UMICH) in May 2014. My PhD thesis lies in the area of multi-terminal information theory. During my doctoral studies I taught graduate level courses both as a student instructor and as a full instructor. In particular, I taught the graduate Probability and Random Processes (EECS501) course offered by the Dept. of EECS, UMICH as a full instructor during the Winter 2014 semester (Jan - Apr 2014).
Briefly, as part of my doctoral thesis, I addressed the long standing problems of characterizing capacity regions of fundamental multi-terminal scenarios such as broadcast and interference networks. I have derived new inner bounds to the capacity regions of these networks with three or more users that strictly improve upon the current known largest inner bounds. The current known largest inner bounds have remained so for over three decades indicating the difficulty of the problems addressed. Please find a detailed description of my findings here.
Through the following links, you can access my preprints and publications, a description of my doctoral thesis and a list of graduate level courses that I have studied through.
My doctoral thesis lies in the area of multi-terminal information theory. In particular, I consider the problem of characterizing inner bounds to the capacity region of multi-terminal communication scenarios such as broadcast and interference channels with three or more users. Employing the ensemble of codes possessing algebraic closure properties, I have derived achievable rate regions for four multi-terminal communication scenarios (including three user broadcast and interference channels) that are strictly larger than the current known largest achievable rate regions. A detailed illustration of my findings can be found here.
I taught the graduate course in Probability offered by the Dept. of EECS as Graduate Student Instructor in Winter 2012 and Winter 2013 terms. In the Winter 2014 term, I taught this course EECS 501 as a full instructor!!!! To teach a graduate course in probability at Univ. of Michigan to Doctoral students immediately after my PhD was a wonderful experience. This provided me with an excellent opportunity to hone my teaching skills.
Email : arunpr[AT]purdue.edu, sieve out the CAPITAL letters!!
Office : Rm 216, Felix Haas Hall, 250 N University St, West Lafayette, IN 47906