DISCLAIMER
These publications are for educational purposes only. They
are provided "as is." Neither Klimasauskas Group, nor Advanced Technology
for Developers, nor its contributors and advisors make any claims for the
content, accuracy, correct operation, fitness, or suitability of the material.
Individuals who download articles and software are specifically prohibited
from using code or programs from any of these publications in any way which
could directly or indirectly cause financial or physical loss, harm or damage
to persons or property. The downloading of any of the material on this web
site does not constitute a license to use. Use at your own risk.
If you do download one or more articles. Please let us know
through the feedback
page. Thank you!
Exposing the Hidden Layer
Susan Garavaglia, Healthcare Analytics Executive, Integrated Therapeutics
Group
Artificial Neural Networks typically contain one input and one output
layer of neural processing elements and one or more intermediate layers
called hidden layers. The purpose of these hidden layers is to encode
an internal representation of the underlying relationship between the
input and output vectors. Hidden layers provide the means for generating
additional connection weights that adapt to some optimal set of values
that form the parameter estimators of a non-linear function. Three different
forms of the hidden layer are discussed, using the same data for each.
Re-print of article published June 1992. Download
170KB ZIP file now.
Taking Backprop to the Extreme: A Master Shares His Secrets
Casey Klimasauskas, Senior Partner, Klimasauskas Group
There are a number of techniques for improving the performance of back-propagation
neural networks based on an understanding of the underlying algorithms.
This article highlights some key concepts and shows their practical
implication in terms of what you can expect from a back-propagation
network. It concludes with a series of simple practical techniques that
can be implemented in many of the popular analytics packages. Published:
PC AI Magazine, January/February
2002. By special arrangement with PC AI Magazine, you can download
a copy of this article (File Size: 6.7MB) for personal use. This
article is Copyright © 2002, PC AI Magazine. All Rights Reserved.
Back to Publication Index |
