The numerical Learning Library intends to provide a wide range of machine learning algorithms. It is a generic and efficient C++ library with its full integrated framework (preprocessing, feature extraction and transformation, validation).
And much more soon…
NLL doesn’t have any dependencies and is mainly composed of header files making it very easy to integrate.
NLL automates a lot of things: parameters of the classifiers, feature selection, preprocessing can all be automatically optimized by different kind of optimizers.
http://sourceforge.net/projects/nll/ by Ludovic Sibille