|
Activity Number:
|
383
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #307375 |
|
Title:
|
Initializing Optimization Partition Algorithms
|
|
Author(s):
|
Ranjan Maitra*+
|
|
Companies:
|
Iowa State University
|
|
Address:
|
Department of Statistics, Ames, IA, 50011-1210,
|
|
Keywords:
|
clustering ; k-means ; E-M algorithm ; singular value decomposition
|
|
Abstract:
|
Clustering datasets is a challenging problem in general, but needed in a wide array of applications. A large number of approaches exist, most of which can be broadly grouped into either the optimization partitioning or the hierarchical clustering class of algorithms. Common examples of the partitioning approaches are the iterative k-means and the expectation-maximization (EM) algorithms. Such algorithms are sub-optimal for multi-dimensional data and find local optima in the vicinity of their initialization. I propose a staged approach for finding starting values. Results on test experiments indicate excellent performance. Applications to clustering mercury emissions data and bioinformatics are presented.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2006 program |
