|
Activity Number:
|
311
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #305318 |
|
Title:
|
Dimension Reduction of Large Datasets in the Atmospheric Sciences
|
|
Author(s):
|
Barbara A. Bailey*+
|
|
Companies:
|
University of Colorado
|
|
Address:
|
Department of Mathematical Sciences, Denver, CO, 80217-3364,
|
|
Keywords:
|
nonlinear PCA ; neural networks
|
|
Abstract:
|
Principal component analysis (PCA) is a multivariate statistical method widely used to reduce the dimensionality of large fields of atmospheric data. A nonlinear extension of the PCA, denoted nonlinear principal component analysis (NLPCA), can explain more of the variance and extract nonlinear features. A feed-forward neural network is used as a flexible nonlinear model to allow nonlinear mappings, whereas PCA only allows for linear mappings. Model selection and statistical properties of the neural network model parameters are investigated. The technique is applied to the noisy Lorenz system and atmospheric science data. Visualization of the results of the NLPCA is discussed.
|
- 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 |
