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

 

Designing Neural Networks

   *   Selecting the Type of Network

 

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.

 

=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=

What’s New and News?

 

Redefining The Learning of Neural Networks

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

 

Neural Network Course Starts Soon

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

This Issue: Selecting the Type of Network

 

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 

 

=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=

Comments or Suggestions?

 

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!