Here we offer resources for modeling cognition using PDP models. To get started with your own neural network, we recommend the PDPTool software and associated documentation, including the PDP Handbook, described below. An alternative is the lens simulator by Doug Rohde. The original PDP software, developed in the late 1980's using a simple character-based user interface, is also still available, although it is no longer maintained or extended. We also offer a copy of the original C code used to implement the TRACE model of speech perception.
PDPTool is a graphical neural network simulator for Matlab that implements many of the most commonly used PDP network types. These include several of the network types described in the PDP volumes (Parallel Distributed Processing: Explorations in the microstructure of cognition by David E. Rumelhart, James L. McClelland, and the PDP research group, 1986), and also several more recent types of networks, including simple recurrent networks, continuous-time recurrent back propagation networks, and Kohonen networks. The simulator enables users to replicate classic simulations and also to create and run their own neural network models that rely on these types of networks. Read an introductory article about PDPTool from the Matlab Digest.
The standard version of the PDPTool software suite runs in a Matlab environment. We recommend Matlab version 7.2 r2006a or higher. To purchase a Matlab license or update your version of Matlab, visit The Mathworks at http://www.mathworks.com. Supported operating systems are Windows NT/XP (SP2 and above), Windows 7, and Apple OSX.
A beta release of a stand-alone version of the sortware is also available for use on Windows, OSX, and Linux operating systems. Instructions for downloading and using the stand-alone version are available in an Appendix of the PDP Handbook, here.
Install PDPTool using the following steps:
- Download the archive file pdptool.zip to your desktop.
- Read the Release Notes for known problems and changes relative to earlier versions.
- Extract the archived files into a new folder called "pdptool".
- Start Matlab.
- Within Matlab, set your path variable to point to PDPTool using the following steps:
- From the File menu, select Set path. A dialog box opens.
- Click the "Add with subfolders" button. A directory browser window opens.
- Locate the folder called "pdptool". Select it and click OK.
- Click the "Save" button on the dialog box to save the path for future sessions.
- Click the "Close" button.
- Set your command history preferences using the following steps:
- From the File menu, select "Preferences". A dialog box opens.
- Select "Command History" from the list of options on the left. This displays the current command history settings.
- In the "Saving" section of the history settings, select "Save after [n] commands", where [n] is a numerical field.
- Change [n] to 1.
- Click "OK".
- At the Matlab command prompt, type "pdp" to start the program.
Requests for help and bug reports can be sent to email@example.com.
The PDP Handbook, now in its second edition, is a free online text that introduces the concepts of parallel distributed processing and neural network models, and provides information about using the PDPTool suite of MATLAB based PDP models.
A tutorial introduction to the PDPTool software can be found in the PDP Handbook. The PDPTool user's guide is a useful adjunct to the handbook, providing an overview of the PDPTool software and its functionality. It also contains descriptions of the application's menus, dialog boxes, and display windows. This is useful as backup documentation, particularly for those wanting to use the PDPTool for systematic research and/or wishing to extend the functionality of the software.
Download it here.
Install the original PDP software using the following steps:
- Download the linked archive file to your desktop.
- Extract the archived files into a new folder named "pdp". Files that should appear include ARCE.COM, COLEX.EXE, and folders for PA, BP, CS, IAC, AA, IA, CL, UTILS, and SRC.
- Refer to the original PDP Handbook for usage instructions.
Procedures for running the software are described in detail in the original PDP Handbook. We regret that we cannot help with recompilation. If you have access to a C compiler, you may be able to make changes and recompile the software for your use.
J. L. McClelland and D. E. Rumelhart's Explorations in Parallel Distributed Processing: A handbook of models, programs, and exercises (Cambridge, MA: MIT Press) is now available online in PDF format.
Access it here.
The TRACE model of speech perception, as used for the simulations reported in McClelland and Elman (Cognitive Psychology, 1986) is available here. Users should be aware that the code is undocumented and uses a very primative ascii text-based interface. Another implementation of TRACE is provided by James Magnuson's laboratory at their jTRACE webpage.
Lens ("the light, efficient network simulator") is an efficient, yet flexible, neural network simulator that runs on a variety of platforms. It can handle large, complex simulations, but is also reasonably easy for novices to operate. The current version was maintained for several years at Carnegie Mellon's Center for the Neural Basis of Cognition, and is now maintained by the Stanford Center for Mind, Brain and Computation.
Access the Lens Manual here.
This project documents a variety of modern neural simulators along a number of useful dimensions. Access it here.
This page contains brief introductions to computational methods that are often useful in the study of human cognition. Access it here.