Greetings
from NeuroDimension!
The World Leader in Neural Network Software
This issue of the NeuroDimension
newsletter highlights significant features of our products and provides you
with updated news and information.
In this issue you’ll find:
What’s New
and News?
* Neural Network Course
Scheduled
* CSW v4 Simplifies Embedding
Neural Networks into your Application
* Using AIC to Determine the
Best Network Size
Note: You are
receiving this newsletter because you requested to stay informed concerning new
developments at NeuroDimension. If you would like to stop receiving these
newsletters, please see the bottom of this newsletter for removal instructions.
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
The next neural network course has been scheduled for April
29 - May 3, 2002 at the Grosvenor Resort, located in the Walt Disney World
Resort in Orlando, Florida. A 10% discount is available for early registration,
so be sure to register today!
Our course format allows both novice and advanced users to
find a suitable course. Offered courses include: "Introduction to
NeuroSolutions", "Fundamentals of Neural Networks and
NeuroSolutions", and "Advanced NeuroSolutions". The courses
include a copy of our interactive book, Neural and Adaptive Systems:
Fundamentals Through Simulations.
For details on this new offering, or to sign-up from the
Internet, see http://www.nd.com/course/may_2002.htm
For general ND course information, see http://www.nd.com/course
For more information and samples of the interactive book,
see http://www.nd.com/products/nsbook.htm
NeuroDimension has a newsletter dedicated to news and tips
especially for TradingSolutions users. If you are interested in using neural
networks for trading and you are only receiving this newsletter, then you are
only getting half the story. Stop by our recently redesigned TradingSolutions
website, take the product tour, and sign up for the newsletter to stay informed
about the latest developments at http://www.tradingsolutions.com
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
For those of you not familiar with the Custom Solution
Wizard (CSW), it is an add-on product for NeuroSolutions that allows you to
automatically create and compile a dynamic link library (DLL) for virtually any
neural network that can be designed within NeuroSolutions. These DLLs respond to a simple protocol that
makes it very easy to embed a neural network model into your own
application. Currently, the programming
tools and other software that can be used with Custom Solution Wizard DLLs
include: Visual Basic, Visual C++,
Microsoft Excel, Microsoft Access, ASP, and TradingSolutions.
When using the Custom Solution Wizard to generate a neural
network DLL, you have the option of also creating a project shell for any of
the project types listed above. This feature was available in the previous
version of the Custom Solution Wizard, but it has been significantly improved
in version 4. Once generated, the project shells now function without any
modification necessary. The training data from the breadboard used to create
the DLL is automatically extracted and configured for use within the project
shell. The result is a fully working neural network application implemented in
the language of your choice. This application can serve as a starting point for
your application or you can cut out the code and use it in an existing application.
Embedding the power of neural networks into your application could not be any
easier.
For more information about the Custom Solution Wizard, see http://www.nd.com/products/csw.htm
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
Many people new to neural networks will often try use a much
larger neural network than is needed for their problem. Generally speaking, if
you add more processing elements to the hidden layer(s), the network will
produce a lower error on the training set. The problem is that when you have
too many free parameters (i.e., network weights), the system is
underconstrained and test set performance may be very poor.
The Akaike information criterion (AIC) is a
performance measure that can be used to help determine the best network size
for your problem. The AIC measure rewards a network with low MSE but penalizes
networks with a large number of weights. For problems with a limited amount of
data, this measurement can be used to determine the network that is least
likely to over-fit the data. Thus, you may be able to avoid using the cross
validation set and instead use all of the available data for training.
The formula for the AIC is a follows:
min AIC(k) = N ln J(k) + 2k
k
where J(k) is the
MSE in the training set, k is the
number of free parameters of the model, and N
is the number of exemplars in the training set. The idea is that the number of
hidden processing elements and/or inputs should be adjusted in order to
minimize this value to find the best compromise between network size and
training set error. As a practical point, since the MSE is generally less than
1, the natural logarithm will create AICs that are normally negative. The
largest negative number is the minimum AIC and the best network.
The AIC is one of the performance measures added to the
Criterion component with the release of NeuroSolutions 4.0. To create a neural
network that displays the performance measures, click the “General” switch on
the Probe Configuration panel of the NeuralBuilder. To display these measures
on an existing breadboard, simply stamp a MatrixViewer probe on top of the
Criterion component and then change the access point of the probe to the
“Performance Measures”. Make sure to double-click the probe icon to display the
probe window.
For more information on the AIC, see section 5.6.3
“Penalizing the Training Error” in our interactive book, Neural and Adaptive
Systems: Fundamentals Through Simulations.
For more information and samples of the interactive book,
see http://www.nd.com/products/nsbook.htm
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
We appreciate your feedback! Please send us your comments or
suggestions concerning this newsletter, our web site, or part of the
NeuroDimension product line. Write to us at: feedback@nd.com
Have questions about NeuroDimension products or training
services? Send your questions to: info@nd.com
Some mail readers may display an attachment at the bottom of
this newsletter. This is typically caused when the mail reader is set to
display text only. The original, formatted version is sometimes included as an
attachment.
This issue and previous issues of this newsletter are
available on the NeuroDimension web site at: http://www.nd.com/mailinglist.htm
If you would prefer not to receive these newsletters or
subsequent product updates from NeuroDimension, please reply to this letter
with the subject heading changed to the word REMOVE.
Thank you again for your support of NeuroDimension products!