More Structured Sparsity

Saturday, June 22, 2013 - 11:00am - 12:30pm
Lind 305
Julien Mairal (INRIA )
In this lecture, we will go beyond Friday's course on structured sparsity, and
consider more complex models. Recently, a large amount of research in
statistics and signal processing has been devoted to developing structured
sparse regularization functions. The goal is to encode some a priori knowledge
about an estimation problem in the regularization, in order to obtain better
prediction or better interpretability. Unfortunately, the literature on the
topic is is vast, and significantly different approaches are now refered to as
structured sparsity. We will present some aspects of this literature, and
focus particularly on convex approaches for structured sparse estimation.

[1] F. Bach, R. Jenatton, J. Mairal and G. Obozinski. Structured Sparsity
through Convex Optimization. Statistical Science. 27(4). 2012
MSC Code: