[Colloquium] Reminder: 1/30 TTIC Colloquium: Constantine Caramanis, UT Austin

[Mary Marre via Colloquium]

When:     Monday, January 30th at 11:00 a.m.

Where:    TTIC, 6045 S. Kenwood Avenue, 5th Floor, Room 526

Who:       Constantine Caramanis, UT Austin


Title: Mixed regression: Fast Algorithms, Optimal Rates, and Multiple
Components

Abstract: Mixture models represent the superposition of statistical
processes, and are natural in machine learning and statistics. We consider
the basic problem of mixed regression. In the worst-case, the problem is
hard. But despite the prevalence of mixture models in practice, even in the
average case, little is known in the realm of efficient algorithms with
strong statistical guarantees.

We provide a convex formulation for this problem, and derive matching
minimax lower bounds. An interesting phenomenon our results show, is that
the minimax rates of convergence are different at low and high SNRs --
something that does not happen in standard regression. On the algorithmic
side, we provide conditions that guarantee convergence of fast EM-like
algorithms in low and high dimensions, and give the first algorithm that is
polynomial in the number of components, with sample complexity scaling
linearly with dimension.


Host: Nathan Srebro <nati at ttic.edu>

For more information on the colloquium series or to subscribe to the
mailing list, please see http://www.ttic.edu/colloquium.php



Mary C. Marre
Administrative Assistant
*Toyota Technological Institute*
*6045 S. Kenwood Avenue*
*Room 504*
*Chicago, IL  60637*
*p:(773) 834-1757*
*f: (773) 357-6970*
*mmarre at ttic.edu <mmarre at ttic.edu>*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20170129/be004671/attachment.html>