On June 2nd I gave a presentation to the Gateway Java Users Group that presented a brief introduction to the Encog AI Framework. I've posted the slides online here for any who wish to view them.

To better test out Encog 2.0 on large training sets I am beginning to conduct research for predicting trends in the S&P 500 index. I am starting out simple just using historic price information to predict future trends. I will post this research, as it progresses, to the following URL:

http://www.heatonresearch.com/market/

There will also be forum posts as this progresses. The first experiment, which was just completed attempted to determine if some companies on the S&P were more predictable than others. You can see a chart of this training here:

The beta version of Encog 2.0 for Java has been released. This is a major update to Encog, and includes many features. You can download Encog 2.0 from http://www.encog.org. I do not believe there are many bugs remaining. If you notice anything, please let me know. Some of the Encog 2.0 features are listed here:

Neural Network Architectures

• Feedforward Neural Network (Perceptron)
• Hopfield Neural Network
• Self Organizing Map (Kohonen)
• Radial Basis Function Network
• Elman Recurrent Neural Network
• Jordan Recurrent Neural Network

Here is the official link from the Gateway Java users group on the Encog presentation that I will be giving. If you are in the St. Louis area, and are interested in Encog, this presentation will give an intro to Encog. I plan to post the material here once the presentation is complete.

http://www.gatewayjug.org/2009/05/june-2nd-meeting-implementing-a-java-n...

The Java docs for Encog 2.0 have been posted. They are about 95% complete. Everything has documentation, I just plan to expand coverage on some of the class definitions. You can see the documentation here.

http://www.heatonresearch.com/javadoc/encog-2.0/index.html

Encog 2.0 beta will be released soon, for now you can download Encog 2.0 from the Heaton Research build server.

http://build.heatonresearch.com/

The C# side is still probably a few months away. Unfortunately, the C# side got fairly out of sync for this upgrade. And it is a major upgrade to Encog.

Encog 2.0 provides a much more flexible structure for representing and training neural networks. A feature list is presented here. Beta versions of Encog will go out in a few days for Java. DotNet, probably in a few months.

A new build server has been added. This computer is responsible for all of the Encog builds. Currently it only handles the Java builds, but soon will handle the C#/DotNet builds as well. We have Cruise Control running on the server you can see it here.

http://build.heatonresearch.com/

If you would like to download the absolute latest build of Encog, this is the place. Of course the latest build is never as stable as our production releases. Currently, we are working on Encog 2.0, which is pretty stable. Defiantly beta quality. You can download the latest from the above page.

On Tuesday, June 2, 2009 I will do a presentation on Encog at the Gateway Java Users Group in St. Louis, MO at 6:00 PM. The information on this presentation can be found here. I will post the material for this presentation online after I complete it.

Implementing a Java Neural Network with the Encog Framework

In this article I am going to present a brief introduction to the Groovy programming language. Groovy can be very beneficial to the artificial intelligence programmer. When training a neural network for pattern recognition it is often necessary to pre-process the data considerably before training the neural network. This is particularly true for temporal pattern recognition. Additionally control and training sets must be established, often by statistical analysis. Groovy allows you to very quickly create such code in a scripted way, without all of the ceremony of regular Java.

One of my goals for Encog past 2.0 is to get more developers than just me working on it. There have been some contributions by others, but mostly, Encog 2.0 has been my own creation, every line of source code. I have a particular need for C# help right now. Encog 2.0 for Java is nearly done, maybe 80% at this point. But many classes have been stabilized. Encog 2.0 was a very large scale change from 1.x. 1.x versions mirrored the book. 2.x gets quite a bit more advanced, and lays the groundwork for me to write more advanced neural network books in Java and C#.