[Colloquium] 8/7 TTIC Colloquium: Aravindan Vijayaraghavan, Northwestern University

[Mary Marre via Colloquium]

When:     Monday, August 7th at 11:00 a.m.

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

Who:       Aravindan Vijayaraghavan, Northwestern University


Title:       Learning Mixtures of Well-Separated Gaussians

Abstract:
Mixtures of spherical Gaussians are among the most widely used statistical
models for clustering. We will consider the problem of efficiently learning
a mixture of a large number of spherical Gaussian components, when the
components of the mixture are well-separated. Here, we are given samples
from a mixture of k spherical Gaussians with parameters corresponding to
the means, variances and mixing weights for each of the components. The
goal is to estimate the parameters up to accuracy \delta using \poly(k,d,
1/\delta) samples.

There is a significant gap between the best known upper bounds and lower
bounds in the amount of separation required for efficiently learning a
mixture of k spherical Gaussians. We will consider the following question:
what is the minimum order of separation needed for learning a mixture of
spherical Gaussians with polynomial samples?

In this talk, I will give a new iterative algorithm for learning mixtures
of k spherical Gaussians, and give sample complexity lower bounds that
together essentially characterize the optimal order of separation between
components that is needed to learn a mixture of k spherical Gaussians with
polynomial samples.

This is based on joint work with Oded Regev.



Host:  <yury at ttic.edu>Yury Makarychev <yury 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/20170731/a236f5ec/attachment.html>