Copyright 1999 - 2003 Steve R. Howell

 NEW! A short SRNEngine Tutorial is now available!

Welcome to the home page of the SRNEngine Neural Network Simulator. SRNEngine is a Microsoft Windows application that makes it easy for the non-programmer to design, run, and analyze neural network simulations. As the name might suggest, SRNEngine is specialized for the Simple Recurrent Network (SRN) architecture, although it will run any feed-forward backpropagation net as well. In addition, many of its analysis features are designed for models of human language or other text processing, although it also has some special features for back-propagation categorization networks.

One of the main advantages of SRNEngine is the way you design networks using it. By simply dragging layers from a toolbar to the WorkSpace, you define input, hidden, context and output layers. Then you drag the layers onto each other to make connections (e.g. drag input to hidden to create a set of input to hidden weights). Once all sets of layers and connections are created, clicking on any layer or connection will display a list of parameters and input/output files that need to be specified. Do that, save the network, go back to the main control panel, and you can run your network! Just like that!

This simulator has proved especially useful with students new to neural networks, especially non-programmers.

NOTE THAT USER DOCUMENTATION, AND CONFIGURATION GUIDES FOR THE DISTRIBUTED COMPUTING SETUP, ARE IN PROGRESS AS OF AUGUST 22, 2003. WE EXPECT TO MAKE THEM AVAILABLE IN A FEW WEEKS. In the meantime, tooltips help is enabled in the program and may assist you to operate it. In addition, a brief overview of the operation of the program is in review for publication, and should be available to you if you are viewing this web site!

SRNEngine has an installation program. Either download the installation files as a single zipped file and then unzip it, or if you prefer, download the three files below separately, and save them to a folder on your hard drive. Be careful that your browser does not save setup.lst as setup.lst.txt, since this is often the default and will cause problems with the installation program.

setup.exe

SETUP.LST

SRNEngine.CAB

Then click on the setup.exe file and it will run the installation, installing to C:\SRNEngine.

However, the version of SRNEngine.exe that is in that installation package is an older one (v1.797). If you plan to install only the background screensaver version of SRNEngine, don't worry. The program will auto-upgrade itself from our FTP site when it first runs.

However, if you plan to use the program directly, you'll want to upgrade to the latest version (v2.017). Click here and save the file into c:\srnengine. Overwrite the existing SRNengine.exe if necessary.

By default, the SRNEngine is a client installation, which sets up a workstation to do unattended simulation runs while the screensaver is activated. We are currently modifying the installation program to both include the most recent exe file and to allow the user to choose whether to install in ScreenSaver or Normal mode. For now, to switch SRNEngine to normal operation do the following after completion of the installation:

(right under the line that says "Mode of Operation (Background (1) or Normal (0)):"

Save the file. Now clicking on the SRNEngine icon will launch the program in Normal mode instead of ScreenSaver mode. Changing that line back to 1 will switch to screensaver mode again. A computer can only use SRNEngine in one mode at a time.

If you are planning to use SRNEngine just on your own PC, you should be able to proceed from here. However, if you want to take advantage of all of the spare PC cycles in your building or department by installing the SRNEngine on those client PC's in screensaver mode, you'll need a little help setting up the configuration files to point to your own departmental FTP master server instead of our central FTP master server. Otherwise, every screensaver you install will be running OUR simulations instead of yours! E-mail Steve R. Howell for assistance with this process.