actually if training data set is too large , u can selectively use the training samples to be the centers of RBFN...but in case of PNN , all the training samples become the centers..So if data set is less,u can select all of them as centers and go for PNN..
I am working on an application using Digital Mammograms for Automatic Detection of Cancer. I am on Feature Extraction Phase of my app, I am planning to use neural network in that app.