We show that, given data from a mixture of k well-separated spherical Gaussians in R
d
, a simple
two-round variant of EM will, with high probability, learn the parameters of the Gaussians to nearoptimal precision, if the dimension is high (d lnk). We relate this to previous theoretical and
empirical work on the EM algorithm
%0 Journal Article
%1 DBLP:journals/jmlr/DasguptaS07
%A Dasgupta, Sanjoy
%A Schulman, Leonard J.
%D 2007
%J Journal of Machine Learning Research
%K EM angular-central-Gaussian mixture sphere
%P 203-226
%T A Probabilistic Analysis of EM for Mixtures of Separated,
Spherical Gaussians
%V 8
%X We show that, given data from a mixture of k well-separated spherical Gaussians in R
d
, a simple
two-round variant of EM will, with high probability, learn the parameters of the Gaussians to nearoptimal precision, if the dimension is high (d lnk). We relate this to previous theoretical and
empirical work on the EM algorithm
@article{DBLP:journals/jmlr/DasguptaS07,
abstract = {
We show that, given data from a mixture of k well-separated spherical Gaussians in R
d
, a simple
two-round variant of EM will, with high probability, learn the parameters of the Gaussians to nearoptimal precision, if the dimension is high (d lnk). We relate this to previous theoretical and
empirical work on the EM algorithm},
added-at = {2011-02-13T19:52:13.000+0100},
author = {Dasgupta, Sanjoy and Schulman, Leonard J.},
bibsource = {DBLP, http://dblp.uni-trier.de},
biburl = {https://www.bibsonomy.org/bibtex/207357fec6aadcf3a45b8a56247762750/ytyoun},
ee = {http://www.jmlr.org/papers/v8/dasgupta07a.html},
interhash = {abc9299500a23f06d86f01a640ad2664},
intrahash = {07357fec6aadcf3a45b8a56247762750},
journal = {Journal of Machine Learning Research},
keywords = {EM angular-central-Gaussian mixture sphere},
pages = {203-226},
timestamp = {2011-10-01T07:39:15.000+0200},
title = {A Probabilistic Analysis of EM for Mixtures of Separated,
Spherical Gaussians},
volume = 8,
year = 2007
}