Feature Space Mapping (FSM) model is based on a network explicitly
modeling probability distribution of the input/output data
vectors. New theoretical developments of this model and results of
applications to several classification problems are presented.