Princeton University

School of Engineering & Applied Science

Universal compressed sensing

Shirin Jalali, Bell Laboratories
E-Quad, B205
Thursday, November 12, 2015 - 4:30pm to 5:30pm

Abstract: Compressed sensing (CS) has made substantial contributions to many data acquisition systems, such as MRI, by dramatically decreasing their sampling times. CS algorithms are designed to recover high-dimensional structured signals from their undersampled measurements. Recovery algorithms are commonly developed based on the properties of signals with a specific type of structure, such as sparsity or low-rankness. However, the massiveness and the diversity of the data we are collecting call for efficient algorithms that require no or very little prior knowledge about the data generation mechanism. In information theory, algorithms that are not tailored for a specific source model and yet efficiently perform their desired tasks are called "universal". In this talk, I will focus on the problem of developing implementable algorithms for universal compressed sensing of stochastic processes. I will talk about the minimum entropy pursuit (MEP) algorithm, which can reliably and robustly recover any Markov process of any order from sufficient number of randomized linear measurements. MEP requires no prior information about the distribution of the source. As I will discuss, the number of measurements required by MEP is connected to the information dimension of the source process, which is a generalization of Renyi's information dimension.
Bio: Shirin Jalali is a researcher in information sciences at Bell Laboratories in Murray Hill, New Jersey. She received her B.Sc. and M.Sc. in electrical engineering from Sharif University of Technology, and her M.Sc. in statistics and Ph.D. in electrical engineering from Stanford University. Her research interests are in the areas of information theory, communications and statistical signal processing.
Before joining Bell Laboratories, she held positions at the department of electrical engineering at Princeton University, the department of electrical engineering at NYU, and the Center for Mathematics of Information at Caltech.