An investigation of computational and informational limits in Gaussian mixture clustering

Authors:
Nathan Srebro;Gregory Shakhnarovich;Sam Roweis
Affiliations:
University of Toronto, Toronto, Ontario, Canada;Brown University, Providence, Rhode Island;University of Toronto, Toronto, Ontario, Canada
Venue:
ICML '06 Proceedings of the 23rd international conference on Machine learning
Year:
2006

Citing 7
Cited 1

Learning mixtures of arbitrary gaussians

STOC '01 Proceedings of the thirty-third annual ACM symposium on Theory of computing
A Two-Round Variant of EM for Gaussian Mixtures

UAI '00 Proceedings of the 16th Conference on Uncertainty in Artificial Intelligence
Learning Mixtures of Gaussians

FOCS '99 Proceedings of the 40th Annual Symposium on Foundations of Computer Science
A spectral algorithm for learning mixture models

Journal of Computer and System Sciences - Special issue on FOCS 2002
A Simple Linear Time (1+ ") -Approximation Algorithm for k-Means Clustering in Any Dimensions

FOCS '04 Proceedings of the 45th Annual IEEE Symposium on Foundations of Computer Science
The spectral method for general mixture models

COLT'05 Proceedings of the 18th annual conference on Learning Theory
On spectral learning of mixtures of distributions

COLT'05 Proceedings of the 18th annual conference on Learning Theory

Are there local maxima in the infinite-sample likelihood of Gaussian mixture estimation?

COLT'07 Proceedings of the 20th annual conference on Learning theory

Quantified Score

Hi-index	0.02

Visualization

Abstract

We investigate under what conditions clustering by learning a mixture of spherical Gaussians is (a) computationally tractable; and (b) statistically possible. We show that using principal component projection greatly aids in recovering the clustering using EM; present empirical evidence that even using such a projection, there is still a large gap between the number of samples needed to recover the clustering using EM, and the number of samples needed without computational restrictions; and characterize the regime in which such a gap exists.