Note to users. If you're seeing this message, it means that your browser cannot find this page's style/presentation instructions -- or possibly that you are using a browser that does not support current Web standards. Find out more about why this message is appearing, and what you can do to make your experience of our site the best it can be.

Science 29 September 1995:
Vol. 269. no. 5232, pp. 1860 - 1863
DOI: 10.1126/science.269.5232.1860
Prev | Table of Contents | Next

Articles

Replicator Neural Networks for Universal Optimal Source Coding

Robert Hecht-Nielsen 1

1 HNC Software Inc., 5930 Cornerstone Court, San Diego, CA 92121, USA, and Department of Electrical and Computer Engineering and Institute for Neural Computation, University of California, San Diego, La Jolla, CA 92093, USA.

Replicator neural networks self-organize by using their inputs as desired outputs; they internally form a compressed representation for the input data. A theorem shows that a class of replicator networks can, through the minimization of mean squared reconstruction error (for instance, by training on raw data examples), carry out optimal data compression for arbitrary data vector sources. Data manifolds, a new general model of data sources, are then introduced and a second theorem shows that, in a practically important limiting case, optimal-compression replicator networks operate by creating an essentially unique natural coordinate system for the manifold.

Submitted on January 24, 1995
Accepted on August 2, 1995


THIS ARTICLE HAS BEEN CITED BY OTHER ARTICLES:
Reducing the dimensionality of data with neural networks..
G. E. Hinton and R. R. Salakhutdinov (2006)
Science 313, 504-507
   Abstract »    Full Text »    PDF »
Non-linear PCA: a missing data approach.
M. Scholz, F. Kaplan, C. L. Guy, J. Kopka, and J. Selbig (2005)
Bioinformatics 21, 3887-3895
   Abstract »    Full Text »    PDF »
A Global Geometric Framework for Nonlinear Dimensionality Reduction.
J. B. Tenenbaum, V. d. Silva, and J. C. Langford (2000)
Science 290, 2319-2323
   Abstract »    Full Text »


To Advertise     Find Products

Science. ISSN 0036-8075 (print), 1095-9203 (online)