ABSTRACT
As in many other scientific and engineering disciplines, in the artificial neu ral networks field, the theory (or at least some part o f it) was only established after a number o f practical neural network applications had been implemented. Many practical questions are still open and for the broad range o f engineering tasks the design o f N N , as wel l as the learning procedures and corresponding training pa rameters, are still in the realm o f the empirical art. In this respect the EBP algo rithm is a genuine representative o f the nonlinear optimization schemes.
