A principal distinction of neural network based methods of approximation from conventional methods is a hidden algorithmic function inside, which is defined by a neural network topology and is enough sophisticate constructed to make appropriate fitting process. There is new construction named a...