[Ml-stat-talks] Irina Rish, IBM, CS 402 Wed. 4/7 1:30 PM
Melissa K. Carroll
mkc at CS.Princeton.EDU
Fri Apr 2 14:01:40 EDT 2010
Irina Rish from IBM Research is visiting the CS department on Wednesday, April 7, and giving the following talk at 1:30 PM in CS 402 (4th floor large conference room). If you are interested in meeting with Irina, please let me know.
Title: Sparse Signal Recovery with Exponential-Family Noise
Accurate and efficient recovery of sparse high-dimensional signals from a relatively low number of samples received much
attention in the recent sparse regression and compressed sensing literature, and led to multiple successful applications including,
among many others, medical imaging and computational biology. Typically, the focus of those studies is limited to
to the standard case of linear projections disturbed by Gaussian noise, and the corresponding formulation of the sparse signal
reconstruction problem as an l1-penalized linear regression (LASSO). However, non-Gaussian noise is often a more appropriate
assumption in various application where the target variable is binary, discrete, or non-negative (brain state prediction from fMRI data
being one such example). Herein, we extend the classical results of [Candes, Romberg,Tao] to the more general case of
exponential-family noise that includes Gaussian as a particular case, and yields l1-regularized Generalized Linear Model (GLM) regression problems. We show that, under standard restricted isometry property (RIP) assumptions on the design matrix, l1-minimization can provide stable recovery of a sparse signal in presence of the exponential-family noise, provided that certain sufficient conditions on the noise distribution are satisfied.
Irina Rish is a research staff member at the Computational Biology Department of the IBM T. J. Watson Research Center. She received an M.S. in applied mathematics from Moscow Gubkin Institute, Russia, and a Ph.D. in computer science from the University of California, Irvine. Dr. Rish's primary research interests are in the areas of probabilistic inference, statistical learning, and information theory, and their applications to large-scale data analysis problems in biology and neuroscience. Her current research focuses on applying machine-learning techniques to neuroscience, and particularly on statistical analysis of fMRI data using sparse regression, dimensionality reduction and graphical models. In the past, she has worked on efficient approximations of probabilistic inference in Bayesian networks, probabilistic diagnosis and experiment design, active learning, collaborative prediction, sparse regression and sparse matrix factorization, and their applications to autonomic computing, as a part of the Adventurous Research project on Self-Managing Computer Systems that she lead at IBM Watson (2003-2007). She has over 40 conference and journal publications on the above topics. Dr. Rish taught several machine learning courses at the Electrical Engineering and Computer Science departments of Columbia University as an adjunct professor, and co-organized several machine-learning workshops at ICML, ECML and NIPS conferences.
More information about the Ml-stat-talks