Greetings from NeuroDimension!
Makers of NeuroSolutions, the Neural Network Simulation Environment.
This is the first issue of a new newsletter designed to help you get the most out of NeuroDimension products, as well as keep you informed of what’s new at ND!
In this issue you’ll find:
What’s New?
Designing Neural Networks
NeuroSolutions Tip Box
Did You Know?
Customer Spotlight
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 page for removal instructions.
What’s New?
Customer Forum on Web Site
Our new Customer Forum is now available to current and prospective customers on our web site. This interactive forum is a great place to get support and information on all NeuroDimension products. Our sales and support staffs are there to help answer your questions and you’ll also find it useful to interact with our other customers to find out how they are solving problems using neural networks. Stop by today and take advantage of this free resource at: http://www.nd.com/forum.htm
New Neural Network Courses Being Offered
NeuroDimension is offering several new courses June 7-11, 1999 in Orlando, Florida. Our new overlapping course format allows novice and advanced users to find the perfect course. Offered courses include: "Introduction to Neural Networks and NeuroSolutions", "Advanced Neural Networks and NeuroSolutions", and "NeuroSolutions for Developers".
For details on the June offerings, see: http://www.nd.com/course/june.htm
For general ND course information, see: http://www.nd.com/course/
Patches for Visual C++ 6.0
If you are using Visual C++ 6.0 and NeuroSolutions v3.02 Professional, Developers Lite, or Developers, then you will want to install the following patch to enable you to use the DLL and/or code generation features: http://www.nd.com/support/nsvc_patch.htm
If you are using Visual C++ 6.0 and any level of the Custom Solution Wizard, then you will want to install the following patch to enable you to build DLLs: http://www.nd.com/support/csw_patch.htm
Interactive Book Now In Beta Test
Our interactive book "Neural Systems: Fundamentals Through Simulations" is now in beta test. The book uses NeuroSolutions to actively demonstrate neural and adaptive systems using over 200 interactive examples. The results have been very positive. With over 25 reviews received, the book received an average overall rating of 4.4 (5=excellent, 1=poor). Every respondent said that the interactive examples were helpful in understanding the topics covered in the book. Comments included: "Very good concept, can't get enough!", "It's a book worth reading. It makes the rather complex theory understandable", and "An excellent contribution and a new dimension for workers in this field." The book is scheduled for publication by Wiley and Sons later this year.
For more information and samples of the interactive book, see: http://www.nd.com/products/nsbook.htm
Designing Neural Networks
This Month: Linear versus Nonlinear Outputs
Nonlinear outputs (e.g. tanh axons or sigmoid axons) work best for classification problems. In classification, the desired output is typically 0 or 1 depending on whether the input is in the appropriate class. Nonlinear outputs that saturate (i.e. those that approach a minimum and maximum value outside of a given range) work well here because they limit the output to the desired range. In fact, under certain conditions the output can be considered an a-posteriori probability that the input is in the class. The probability can never be more than one or less than zero.
Linear outputs (e.g. linear axons or bias axons) typically work best for function approximation problems, such as prediction, modeling, or control applications. In these problems the desired output is continuous, which can cause problems when using nonlinear outputs. First, the range of nonlinear outputs is restricted, which limits your ability to extrapolate beyond your input data. Second, the extreme values of your data (e.g. near 1) are treated differently than the other values (because of the nonlinearity). This is typically undesirable.
Remember, these rules of thumb apply only to the output layer. In a neural network, any hidden layer axons should be nonlinear.
NeuroSolutions Tip Box
This Month: Exemplar Weighting
Classification problems often do not have an equal number of training exemplars (samples) for each class. For example, you may have a neural network application that detects the occurrence of cancer from clinical test data. The training data for this problem may contain 99 exemplars classified as non-cancerous for every one exemplar classified as cancerous. A standard neural network would most often train itself to classify all exemplars as non-cancerous so that it would be 99% correct. Since the goal is to detect the existence of cancer, this is a problem.
NeuroSolutions provides a solution to this called exemplar weighting. For the example above, each of the cancerous training exemplars would have 99 times more weight during the backpropagation procedure than the non-cancerous exemplars. This balancing of the training data will most likely result in a system that does a much better job of detecting the cancerous cases.
Exemplar weighting is activated from the "Weighting" page of the inspector for the BackControl component (right-click on the red dials and select "Properties"). Use the Context Help on this inspector page to obtain information on the user interface and the formula used for the default weighting.
Did You Know?
This Month: Custom Batches with NeuroSolutions for Excel
Did you know that NeuroSolutions for Excel allows you add your own custom batches? If you frequently pre-process your data in a particular manner before sending it to a neural network, you can write a custom batch to perform this pre-processing. In fact, custom batches can be written for any of the seven NeuroSolutions for Excel modules.
Custom batches are easy to write. You simply select "New Batch" from the "NeuroSolutions" menu in Excel, select the type of batch you want to create, enter a name, and then click "OK". The Visual Basic Editor will be launched automatically and a new module will be created for your custom batch. The module contains three functions: OnGetName, OnGetDescription, and OnRunBatchProcess. Use the OnGetName function to return a name for your batch, the OnGetDescription function to return a description for you batch, and the OnRunBatchProcess for the implementation of you batch (see the on-line help for more details). After your have written your batch, you can run it from the NeuroSolutions menu by selecting the appropriate NeuroSolutions for Excel module and clicking "Run Batch".
Have a hot tip for using a NeuroDimension product? Send it to us for inclusion in the newsletter or share it with others in the Customer Forum at: http://www.nd.com/forum.htm
Customer Spotlight
This Month: Neural Network Based Power Plant Coal Quality Analysis
University of North Dakota Department of Electrical Engineering and Microbeam Technologies, Inc.
Ash problems in coal-fired power plants result in decreases in efficiency, unscheduled outages, equipment failures, and cleaning. Assessing the potential impact of ash on power plant performance is extremely complex and difficult due to coal variability, the complexity of the ash behavior processes, and changing operating conditions. To predict the impact of ash on power plant performance, the impurities and mineral contents of coal have to be determined. Current coal quality evaluation methods are either inefficient or very expensive and time consuming. A neural network has been developed which quickly determines the impurities and ash forming species in coal. The developed model shows promise and has the potential to save coal-fired utilities millions of dollars in dealing with various coal ash problems.
A complete summary of this application is available in PDF format for downloading from: http://nd.alabanza.com/anonftp/pub/appsum/app-quality.pdf
Want to have your solutions spotlighted? We strongly encourage our customers to send their 1-2 page application summaries to submissions@nd.com so that we may post them on our web site at: http://www.nd.com/appliactionsum.htm. In each newsletter, we’ll spotlight a new solution and include a link for people to get more information.
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
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!