Auto-SOM: recursive parameter estimation for guidance of self-organizing feature maps

Haese, K. and Goodhill, G. J. (2001) Auto-SOM: recursive parameter estimation for guidance of self-organizing feature maps. Neural Computation, 13 3: 595-619. doi:10.1162/089976601300014475

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
UQ109232_OA.pdf Full text (open access) application/pdf 1.08MB 0

Author Haese, K.
Goodhill, G. J.
Title Auto-SOM: recursive parameter estimation for guidance of self-organizing feature maps
Journal name Neural Computation   Check publisher's open access policy
ISSN 0899-7667
Publication date 2001-01-01
Year available 2001
Sub-type Article (original research)
DOI 10.1162/089976601300014475
Open Access Status File (Publisher version)
Volume 13
Issue 3
Start page 595
End page 619
Total pages 25
Place of publication Cambridge, MA, United States
Publisher M I T Press
Language eng
Abstract An important technique for exploratory data analysis is to forma mapping from the high-dimensional data space to a low-dimensional representation space such that neighborhoods are preserved. A popular method for achieving this is Kohonen's self-organizing map (SOM) algorithm. However, in its original form, this requires the user to choose the values of several parameters heuristically to achieve good performance. Here we present the Auto-SOM, an algorithm that estimates the learning parameters during the training of SOMs automatically. The application of Auto-SOM provides the facility to avoid neighborhood violations up to a user-defined degree in either mapping direction. Auto-SOM consists of a Kalman filter implementation of the SOM coupled with a recursive parameter estimation method. The Kalman filter trains the neurons' weights with estimated learning coefficients so as to minimize the variance of the estimation error. The recursive parameter estimation method estimates the width of the neighborhood function by minimizing the prediction error variance of the Kalman filter. In addition, the "topographic function" is incorporated to measure neighborhood violations and prevent the map's converging to configurations with neighborhood violations. It is demonstrated that neighborhoods can be preserved in both mapping directions as desired for dimension-reducing applications. The development of neighborhood-preserving maps and their convergence behavior is demonstrated by three examples accounting for the basic applications of self-organizing feature maps.
Keyword Computer Science, Artificial Intelligence
Energy Functions
Q-Index Code C1
Q-Index Status Provisional Code
Institutional Status Unknown

Document type: Journal Article
Sub-type: Article (original research)
Collection: Queensland Brain Institute Publications
Version Filter Type
Citation counts: TR Web of Science Citation Count  Cited 15 times in Thomson Reuters Web of Science Article | Citations
Scopus Citation Count Cited 16 times in Scopus Article | Citations
Google Scholar Search Google Scholar
Created: Thu, 20 Sep 2007, 02:19:26 EST