A Neural Network for Feature Extraction
Abstract
This paper suggests a statistical framework for the parameter estimation problem associated with unsupervised learning in a neural network, leading to an exploratory projection pursuit network that performs feature extraction, or dimensionality reduction. The search for a possible presence of some unspecified structure in a high dimensional space can be difficult due to the curse of dimensionality problem, namely the inherent sparsity of high dimensional spaces. Due to this problem, uniformly accurate estimations for all smooth functions are not possible in high dimensions with practical sample sizes. Recently, exploratory projection pursuit (PP) has been considered (Jones, 1983) as a potential method for overcoming the curse of dimensionality problem and new algorithms were suggested by Friedman and by Hall. The idea is to find low dimensional projections that provide the most revealing views of the full- dimensional data emphasizing the discovery of nonlinear effects such as clustering. Many of the methods of classical multivariate analysis turn out to be special cases of PP methods. Examples are principal component analysis, factor analysis, and discriminant analysis. The various PP methods differ by the projection index optimized.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1990
- Accession Number
- ADA223059
Entities
People
- Nathan Intrator
Organizations
- Brown University