View this message in a browser.

Neurosolutions Newsletter
Visit neurosolutions.com | Download Evaluation Volume 7, No.3
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.
Getting Started with NeuroSolutions for Matlab

The 15 flexible neural models included with NeuroSolutions for MATLAB are sufficient to solve many different types of problems. Many popular architectures are included, such as Multi-Layer Perceptrons (MLP's), Support Vector Machines (SVM's), and Time-Lag Recurrent Networks (TLRN's). For developers and researchers requiring even more flexibility, NeuroSolutions for MATLAB can be integrated with NeuroSolutions through the Custom Solution Wizard to access virtually unlimited custom neural network architectures.

Building a Simple Neural Network

In NeuroSolutions for Matlab, "Demo 4" allows you to use your own data or the XOR data set included with the software to quickly build a neural network in Matlab. To bring up the demos, simply type "nsDemos" in the Matlab interface or by clicking on the Start menu, NeuroSolutions for Matlab 2.0 and Demos.

As you step through the demo it will display the NeuroSolutions for Matlab code used in generating the appropriate breadboard:

mynet = nsnn;
mynet.numHiddenLayers = 2;
mynet.modelSettings.gradientDescentMethod = 'LevenbergMarquardt';
mynet.outputAndDesired = logical(1);
mynet = nsTrain (mynet, xor_in, xor_desired);

Your simple neural network is ready to use in Matlab! Of course NeuroSolutions for Matlab goes further in-depth including adding your own custom neural networks from NeuroSolutions as described in the next section.

Creating Custom Neural Networks

The Custom Solution Wizard allows you take custom neural networks you have created with NeuroSolutions and generate them as DLL's you can use in other applications. NeuroSolutions 5.0 and the latest version of NeuroSolutions for Matlab v2.0 has the ability to generate DLL's directly for use with NeuroSolutions for MATLAB and Matlab 2007a.

In addition to generating the neural network DLL, the Custom Solution Wizard also automatically loads the input and desired data into MATLAB and prints out the necessary training and testing code on the MATLAB command line. Your custom neural network is ready to use, complete with data.

Deploying Custom Neural Networks in MATLAB

The following three lines of MATLAB code are all it takes to use your generated NeuroSolutions neural network through NeuroSolutions for MATLAB:

mynet = nsnn;
mynet.customDLL = true;
mynet.dllPath = 'C:\My DLLs\myDLL.dll';

Your custom neural network is now ready to use in MATLAB. You can then use it the same way you would use any neural network in NeuroSolutions for MATLAB. For example, to train and test your network, you would use the following commands.

mynet = nsTrain (mynet, myInputData, myDesiredData);
out = nsTest (mynet, myTestingData);

The NeuroSolutions for MATLAB Suite

NeuroSolutions, NeuroSolutions for MATLAB, the Custom Solution Wizard, and NeuroSolutions for Excel are available as a combined package called the NeuroSolutions for MATLAB Suite. This discounted bundle includes all of the product levels necessary to create and generate your own custom neural networks for use in NeuroSolutions for MATLAB.

For more information about NeuroSolutions for MATLAB and the NeuroSolutions for MATLAB Suite, please click the "More Info" button below.