CaltechAUTHORS
  A Caltech Library Service

Correlations in high dimensional or asymmetric data sets: Hebbian neuronal processing

Softky, William R. and Kammen, Daniel M. (1991) Correlations in high dimensional or asymmetric data sets: Hebbian neuronal processing. Neural Networks, 4 (3). pp. 337-347. ISSN 0893-6080. http://resolver.caltech.edu/CaltechAUTHORS:20130816-103228638

Full text is not posted in this repository. Consult Related URLs below.

Use this Persistent URL to link to this item: http://resolver.caltech.edu/CaltechAUTHORS:20130816-103228638

Abstract

The Hebbian neural learning algorithm that implements Principal Component Analysis (PCA) can be extended for the analysis of more realistic forms of neural data by including higher than two-channel correlations and non-Euclidean 1p metrics. Maximizing a dth rank tensor form which correlates d channels is equivalent to raising the exponential order of variance correlation from 2 to d in the algorithm that implements PCA. Simulations suggest that a generalized version of Oja's PCA neuron can detect such a dth order principal component. Arguments from biology and pattern recognition suggest that neural data in general is not symmetric about its mean; performing PCA with an implicit 1l metric rather than the Euclidean metric weights exponentially distributed vectors according to their probability, as does a highly nonlinear Hebb rule. The correlation order d and the 1p metric exponent p were each roughly constant for each of several Hebb rules simulated. High-order correlation analysis may prove increasingly useful as data from large networks of cells engaged in information processing becomes available.


Item Type:Article
Related URLs:
URLURL TypeDescription
http://dx.doi.org/10.1016/0893-6080(91)90070-LDOIArticle
http://www.sciencedirect.com/science/article/pii/089360809190070LPublisherArticle
Additional Information:Copyright © 1991 Published by Elsevier Ltd. Received 7 May 1990; Accepted 30 October 1990; Available online 6 March 2003. D. M. K. is supported by a Weizmann Postdoctoral Fellowship from the Division of Biological Sciences. This work was supported by grants to C. Koch from James S. McDonnell Foundation, the Air Force Office of Scientific Research, and a NSF Presidential Young Investigator Award. W. S. is grateful for suggestions and encouragement from Pierre Baldi. D. K. is grateful for encouragement from Christof Koch. We would also like to thank Fernando Pineda and an anonymous reviewer for comments.
Group:Koch Laboratory, KLAB
Funders:
Funding AgencyGrant Number
Weizmann Postdoctoral Fellowship (Caltech)UNSPECIFIED
James S. McDonnell FoundationUNSPECIFIED
Air Force Office of Scientific ResearchUNSPECIFIED
NSF Presidential Young Investigator AwardUNSPECIFIED
Subject Keywords:Principal component analysis, Hebbian learning, Self-organization, Correlation functions, Multidimensional analysis, Non-Euclidean metrics, Information theory, Asymmetric coding
Record Number:CaltechAUTHORS:20130816-103228638
Persistent URL:http://resolver.caltech.edu/CaltechAUTHORS:20130816-103228638
Official Citation:William R. Softky, Daniel M. Kammen, Correlations in high dimensional or asymmetric data sets: Hebbian neuronal processing, Neural Networks, Volume 4, Issue 3, 1991, Pages 337-347, ISSN 0893-6080, http://dx.doi.org/10.1016/0893-6080(91)90070-L. (http://www.sciencedirect.com/science/article/pii/089360809190070L)
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:40515
Collection:CaltechAUTHORS
Deposited By: KLAB Import
Deposited On:11 Mar 2010 05:57
Last Modified:02 Dec 2013 23:16

Repository Staff Only: item control page