Hello everybody

I implemented the orthogonal least squares algorithm for radial basis functions networks with a single output neuron   - based on the documentation written by Chen, Cowan and Grant.
Now I would like to generalize the algorithm - which now works only for prediction - so that to work for classification too. I mean I want it work now for a RBF network with
more than one neuron on the output layer.
Does anyone have a mathematical specification or an algorithm description for this case?

Thank you
Alina Dana COROIAN
SIEMENS VDO Automotive