 |
 |
 Visit neurosolutions.com | Download Evaluation |
Volume 6, No.9 |
| This is a Neurosolutions Newsletter, which you are receiving because you requested to stay informed about new developments at NeuroDimension. You can also access current and past issues of Neurosolutions Newsletters online. |
|
|
|
| Holiday Sale - Save 20%
NeuroDimension is excited to offer the Happy Holidays Sale! From now through December 30th, all NeuroDimension software can be purchased at 20% off the regular list prices.
|
|
|
This is an excellent time to purchase NeuroSolutions, TradingSolutions, a NeuroSolutions or TradingSolutions add-on, or one of our genetic products!
But don't wait too long - this offer ends December 29, 2006. Use offer code 34211 when placing your order. You can place your order now using our online order entry system at http://www.neurosolutions.com/order/
Note: This offer cannot be used in conjunction with any other special offer or on Bowfort Technologies' Fuzzy Candlesticks add-on for TradingSolutions.
|
|
|
| Neural Network Course a Resounding Success!
We successfully completed another Neural Network and NeuroSolutions Course in early November. We always enjoy getting to know our customers better, especially when it gives us a chance to discuss their applications. Once again, we received excellent reviews from our attendees and have never had anyone tell us they would not recommend our course to others.
|
|
|
Quote from Attendee:
"This is an excellent course for understanding NN (neural network) concepts and applying them on real examples. I think the course is unique of a kind, at least in my search."
The next Neural Network course is being planned for Spring 2006. If you would like more information on our courses, please click the "More Info" button below.
|
|
| NeuroSolutions Tip Box: Using Weight Decay to Improve Generalization (updated)
One of the points emphasized during our neural network course is the importance of generalization. A neural network is not going to be very useful if it “memorizes” the training set and does a poor job of modeling an out-of-sample test set. One cause of this poor generalization is overtraining. Overtraining can be prevented by stopping the training (or at least saving the network weights) when the error of the cross validation set begins to rise.
|
|
|
Another cause of poor generalization is using a neural network that has too many free parameters (weights). A good rule of thumb is to have a system that has only 1 weight for every 5-10 exemplars (input/output pairs) of training data. One way to reduce the number of weights in the neural network is using a technique called weight decay. Weight decay subtracts a small amount from each weight each time it is updated. If the weight is not relevant, it eventually moves at or near zero.
Weight decay is an option within the Step, Momentum, Quickprop and Delta-Bar-Delta components of the GradientSearch toolbar. To activate the algorithm, simply click on the “Decay Weights” checkbox within the Momentum or Step tab of the component's inspector. You will also need to specify the decay rate. Higher decay rates result in the weights decaying faster, causing more of the weights to go to zero. Lower decay rates decrease this effect, with a decay rate of zero being equivalent to the weight decay algorithm being deactivated. If you find that the learning curve jumps around or does not train to a low enough error, then you may have the decay rate set too high.
For more information on weight decay and similar algorithms, see the interactive book, Section 4.7, Network Size and Generalization.
 |
|
| Application Section Enhancements
Ever wonder if neural networks could be applied to your area of expertise? Well look no further than the new Application Section at http://www.nd.com/apps. This new section includes examples of neural network uses, links to papers with related studies, and pointers to relevant examples you can try in NeuroSolutions.
|
|
|
We are always looking for new ways to educate our users on neural network uses so they can make the most of our products. We recently added some "real world" data examples to each section that can run be used in NeuroSolutions and also Customer Interviews that pertain to users using NeuroSolutions in their "real world" studies. For more information on the Application Section, please click on the "More Info" button below.
|
|
| New Web Content Overview
New content is always being added to our websites. Here is some of the latest content of interest to our users:
|
|
|
|
|
|
|
If you would prefer not to receive subsequent product updates or newsletters from NeuroDimension, please feel free to advise us by email and we will be happy to update your details on your behalf.
|
|
