View Image

Walnut is an OpenSource framework for the development of neural network simulations by both beginners and experts. It is a system that is extensible with new paradigms. Neural network architectures and simulations are portable between various platforms (notably Windows and Unix). These objectives are met by having:

* Run-time loadable neural network paradigms (developed in C++ and accessed as shared libraries), such as backpropagation, Kohonen, etc.
* An XML-like, platform-independent file format for saving and loading of architectures, called Neural Network Markup Language or NNML.
* Specification of simulations through platform independent Python scripts (although we also support VBA Basic and other scripts on Windows, while 'scripting' in C++ is also possible).

For the Windows platform, we currently also have a feature-rich browser or shell in which Walnut neural networks can be constructed, visualized, and manipulated. This system is called Nutshell. Walnut and Nutshell are available for free.

At this time several architectures have already been implemented:

* Hopfield
* Backpropagation
* McGregor
* Kohonen
* TraceLink
* Calm

A neural network browser or shell, called Nutshell, has been made that allows students and researchers to visualize, build and manipulate models with the various network paradigms. Simulations can be scripted in Visual Basic (e.g., VBA in Excel), Python and other scripting languages. Nutshell can be downloaded free of charge. At this time there is only a version that operates in MS Windows, but ports to other platforms (Linux) are under development.

Both Nutshell and Walnut are free software, distributed under the terms of the GNUGeneral Public License ('GPL') and GNULesser General Public License ('LGPL'), respectively, with this Copyright.