What is it?
NeuroKit is an object-oriented framework implementing neural networks. Additionally to simply providing implementations of the various algorithms in use today it has a strong emphasis on data import and export facilities. Applying machine learning to a specific problem usually involves the preparation of input data which can be very time consuming.
The main goal of this project is to provide a complete set of classes that allow using neural networks in foreign application with minimum effort. This includes to relieve the programmer of an application from the tedious task of data import.
First of all, GNUstep provides very useful and solid object-oriented framework.
NeuroKit is based on code I have been working with during the last few years. So there already is some code base. The problem is that the code that NeuroKit should be based is spread among several projects implemented in C, C++, Delphi and Objective-C. I am currently in the process of sorting out usable code and reimplement it using the GNUstep libraries.
The major code base of NeuroKit already is implemented in Objective-C, though not using GNUstep. It includes implementations of the Backpropagation algorithm, Self-Organizing Maps and support for some text based data formats.
The main platform used for implementing and testing NeuroKit will be GNUstep running on GNU/Linux (at least for now). But there should be no major problem to get NeuroKit running on any system supported by GNUstep. This includes various UN*X-like operating systems and systems based on Windows NT.