Greetings
from NeuroDimension!
The World Leader in Neural Network Software
This issue of the NeuroDimension
newsletter highlights how NeuroDimension can help you to learn about neural
networks and integrate them into your products. In addition, it includes a
valuable overview of many popular neural network architectures.
In this issue you’ll find:
What’s New
and News?
* Redefining The Learning of
Neural Networks
* Neural Network Course Starts
Soon
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.
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
It has been two and a half years since we first released our
interactive book, Neural and Adaptive Systems: Fundamentals Through
Simulations. During that time, thousands of copies of this innovative book
have been sold to professors, students, companies, and individuals, all wanting
to learn more about neural networks.
A combination of traditional hard cover textbook, hypertext
e-book, and software – the interactive book works directly with an included
version of NeuroSolutions to provide over 200 interactive experiments to
elucidate the fundamentals of neural networks and adaptive systems.
Here are what just a few of its many satisfied readers have
had to say about this innovative approach to learning:
“This has got to be the best textbook I have come across!
The amount of thought and care in producing this book is phenomenal. It goes
into details very thoroughly, and provides plenty of examples which would
easily satisfy the very bright and the very slow. The math in back of all this
is ACTUALLY EXPLAINED PROPERLY! Buy the book - it's worth every cent."
-- Jenni Gyffyn
"Only one word to describe this book, WOW. Neural
nets always seemed to be a mystery. I search the net looking for a way to learn
AI when I crossed NeuroDimension website. Their software caught my attention. I
order the book to get a better understanding. When I received the book and CD,
I was in Awe. This book/online book is so comprehensive and easy to understand.
I cannot wait to implement neural nets in my next applicable system design."
-- Chuck Streb, System Integrator
For more information and samples of the interactive book,
visit the NeuroDimension web site. http://www.nd.com/products/nsbook.htm
Our next series of neural network courses is rapidly
approaching. However, there are still a few seats available. These courses are
scheduled for April 29 - May 3, 2002 at the Grosvenor Resort located in the
Walt Disney World Resort in Orlando, Florida. Contact us immediately if you are
interested in filling one of these last remaining vacancies.
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 course, 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
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
Designing Neural Networks
There are
many different types of neural networks and each one has its own strengths and
weaknesses. Often times, selecting the appropriate architecture can make the
difference between success and failure.
NeuroSolutions
has the flexibility to create virtually any neural network architecture. Beginning users can let the NeuralExpert
wizard automatically select and configure good neural network architectures
for many general applications. Intermediate users can use the NeuralBuilder
wizard to select and configure eleven of the most popular architectures.
Advanced users can then modify these architectures or design their own neural
network topologies from scratch.
This
article will give a brief discussion of the most popular architectures. For
more detailed information about these architectures and others, pick up a copy
of the interactive book, Neural and Adaptive Systems: Fundamentals Through
Simulations or sign up for one of our neural network courses.
Multi-layer
Perceptron (MLP)
The most
popular network is the multi-layer perceptron (MLP), and with good reason. This
is the workhorse of the neural network world and works well on the largest
percentage of problems. The MLP is most adept at solving static problems
(problems which do not depend on information in time, e.g. not time series
prediction). It scales well with the dimensionality of the input and
generalizes well as long as it is appropriately sized. We select the MLP by
default for static classification and function approximation problems. (For
more information on network sizing, see the April 1999 newsletter.)
Radial
Basis Function Network (RBF)
The radial
basis function network (RBF) is similar to the MLP in its broad application and
in general can be applied anywhere that the MLP can. The RBF, however, does not
scale well with the dimensionality of the input (e.g. number of features). The
RBF, however, has a much more local structure than the MLP such that outliers
and other “issues” in one region of the input will not affect the others.
Support
Vector Machine (SVM)
The
support vector machine (SVM) is a new architecture that optimizes the decisions
it makes, rather than the mean squared error. Theoretically, this provides much
better generalization. The two main restrictions on the SVM are that it
currently can only be used for classification (although there are methods being
developed for function approximation) and that its complexity is related to the
number of input exemplars (or data records). Thus, it tends to work better on
small data sets, which is also when MLP’s tend to be weaker.
ANFIS
Neural Fuzzy Architecture
The ANFIS
neural fuzzy architecture is an adaptive system that allows a priori
knowledge (e.g. a rule base) to be built into the system and then optimized
based on the data. Its main advantage is that it is possible to extract
information about the rules used to make a decision.
Time-Lagged
Architectures
The
time-lagged recurrent neural network (TLRN) or time-lagged feedforward neural
network (TLFN) are two names for an MLP with memory. The simplest case is the
time-delayed neural network (TDNN) that is an MLP with a delay line or window
of past inputs presented to the MLP instead of just the current data.
These
architectures allow the power of the MLP to be utilized when solving problems
where temporal information is critical. A typical example of this is time
series prediction where the trend in the data is at least as important as the
current data point in predicting the future. For example, TradingSolutions uses
time-lagged neural networks as its default neural network for predicting and
modeling of stock market data.
Unsupervised
Networks
For data
compression, preprocessing, and data mining you can use unsupervised networks
(or hybrid networks). These networks work without a “desired signal” and
therefore can only extract relationships between the inputs. Unsupervised
networks can be used for clustering, information extraction such as principal
component analysis, visualization and data mining. Self-organizing maps (SOM’s)
are a very powerful tool used for these last two tasks.
Again,
this was only a brief discussion of the most popular architectures. For more
detailed information about these architectures and others, pick up a copy of
the 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
For information about upcoming NeuroDimension neural network
courses, see http://www.nd.com/course
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
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!