Welcome to the Signal and Information Processing Seminar Series at Rutgers!
The SIP Seminar Series at Rutgers University–New Brunswick brings together a diverse group of researchers, both from within and outside Rutgers, on (approximately) a biweekly basis to discuss recent advances in signal and information processing. The term of “Signal and Information Processing” used within the SIP Seminars is rather broad in nature and subsumes signal processing, image processing, statistical inference, machine learning, computer vision, control theory, harmonic analysis, information theory, etc.
Seminar Mailing List: You can subscribe to the SIP Seminars mailing list by sending an email to ECE_SIPfirstname.lastname@example.org with SUBSCRIBE in the subject of the email.
Fall 2018 Seminar Schedule: The SIP Seminars in Fall 2018 will take place on the following dates (typically, Wednesdays) at 2 pm in Room 240 of the Electrical Engineering Building on Busch Campus of Rutgers University–New Brunswick: Sep. 26; Oct. 24; Nov. 7, 21; and Dec. 5.
Fall 2018 SIP Seminars
Dr. David Karpuk
Title: Private Computation
Abstract: Private Information Retrieval (PIR) is the problem of downloading a file from a database without revealing to the database which file is being downloaded. This problem has been studied extensively from an information-theoretic perspective over the past few years, and schemes which achieve the optimal download rate are known for a variety of generalizations of the PIR problem. Private Computation (PC) is a generalization of the PIR problem in which a user wishes to compute an arbitrary function of the database, without revealing the identity of the function. Much less is known about this problem, especially when the function one wishes to compute is not linear. In this talk we will give a survey of the current state-of-the-art in PIR and construct some explicit PC schemes for computing polynomial functions. We will conclude by discussing some very recent progress on the problem of Private Search, in which a user wishes to search a database for a file close to their own file, while hiding the contents of their file.
Biography: David Karpuk received his Ph.D. in Mathematics from the University of Maryland, College Park in 2012. As a Postdoctoral Researcher at Aalto University, Finland from 2012-2017, his research focused on applying algebraic and number-theoretic tools to problems in coding theory and wireless communications. Since Fall 2017 he has been an Assistant Professor at Universidad de los Andes, Bogotá, Colombia, and his current research interests involve privacy and security in distributed computation.
Dr. Alex Dytso
Title: Generalized Gaussian Distributions in Estimation and Information Theory
Abstract: This talk will focus on the family of distributions termed Generalized Gaussian (GG). The family of GG distributions has received considerable attention from the engineering community, due to the flexible parametric form of its probability density function, and used for modeling many physical phenomena. Roughly the talk will consist of four parts. In the first part of the talk, we will consider problems of estimation of GG random variables in the presence of Gaussian noise. Due to the fact that the pdf of GG distributions is not always analytic, the Bayesian Cramer-Rao lower bound in this setting is either too lose or does not exist. We will show a new class of lower bounds that are based on the study of non-Gaussianity and are tight in this setting. The GG data sources are ubiquitous, thus it is important to understand the fundamental limits of lossy compression of GG sources. In the second part of the talk, we will discuss the lossy compression of GG sources. Closed-form expressions for the rate-distortion curves will be given. Along the way, we will show previously unknown properties of the GG distribution. For example, a complete characterization of conditions under which GG random variables are infinitely divisible and self-decomposable will be given. The third part of the talk will focus on communication over channels with additive GG noise. The GG distributions can model impulsive noise environments such as acoustic under-water noise and interference in ultrawideband systems with time-hopping. The final part is an outlook focusing on the open problems and future directions.
Biography: Alex Dytso is currently a Postdoctoral Researcher in the Department of Electrical Engineering at Princeton University working under the supervision of Professor H. Vincent Poor. He received a Ph.D. degree from the Department of Electrical and Computer Engineering at the University of Illinois at Chicago (UIC) under the supervision of Daniela Tuninetti and Natasha Devroye. He received his B. S. degree in 2011 from the University of Illinois at Chicago where he received International Engineering Consortium’s William L. Everitt Student Award of Excellence for outstanding seniors, who have demonstrated an interest in the communications field. His current research topic focuses on multi-user information and estimation theories and their applications to wireless networks.
Title: Distributed Learning Under Byzantine Failures
Abstract: Distributed machine learning algorithms enable learning of models from datasets that are distributed over a network without gathering the data at a centralized location. While efficient distributed algorithms have been developed under the assumption of faultless networks, failures that can render these algorithms nonfunctional occur frequently in the real world. This talk focuses on the problem of Byzantine failures, which are the hardest to safeguard against in distributed algorithms. While Byzantine fault tolerance has a rich history, existing work does not translate into efficient and practical algorithms for high-dimensional distributed learning. In this talk, two variants of an algorithm termed Byzantine-resilient distributed coordinate descent (ByRDiE) are introduced that enable distributed learning in the presence of Byzantine failures. Theoretical analysis and numerical experiments presented highlight the usefulness of ByRDiE for high-dimensional distributed learning in the presence of Byzantine failures.
Biography: Zhixiong Yang received the B.E. degree (with Honors) in electrical engineering from Beijing Jiaotong University, Beijing, China, in 2011. He is currently a Ph.D. student with the Department of Electrical and Computer Engineering, Rutgers University-New Brunswick, NJ, USA. His research interests include distributed optimization methods, robust learning algorithms, and learning under adversarial settings.
Dr. Shirin Jalali
Prof. Mert Gurbuzbalaban