The new modifications of multilayered neurak networks training algorithms in a generalized training plant structure are introduced. The first modification for algorithm "on error backpropagation algorithm through time" is introduced and its sufficient conditions of a training procedure stability are ...