Activity Number:
|
697
|
Type:
|
Contributed
|
Date/Time:
|
Thursday, August 4, 2016 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Statistical Learning and Data Science
|
Abstract #319505
|
View Presentation
|
Title:
|
Sparse Clustering of High-Dimensional Gaussian Mixtures
|
Author(s):
|
Jing Ma* and Tony Cai and Linjun Zhang
|
Companies:
|
University of Pennsylvania and University of Pennsylvania and University of Pennsylvania
|
Keywords:
|
Normal mixtures ;
Unsupervised clustering ;
EM algorithm ;
High-dimensional inference
|
Abstract:
|
This work studies the problem of high dimensional clustering when the observations come from a two-class normal mixture with equal covariances. In particular, we investigate the performance of the EM algorithm in estimating sparse discriminant vectors using regularized maximum likelihood. With an appropriate initialization, our results show that the EM algorithm attains an estimator with faster convergence rate than the commonly used method of moments. We further establish that unsupervised clustering equipped with the above estimator yields superior performance in terms of misclustering rate.
|
Authors who are presenting talks have a * after their name.