The diploma project is about training methods for artificial neural networks (ANN). The problem is to classify dynamical processes by momentum measurements using ANN. The idea is to upgrade neural network weights to time-depended functions. In the paper affine weight functions were studied. By combining backpropagation and speed-gradient method the new algorithm to train ANN was obtained. In a case of insufficient time there are no good application example.  My opinion is “good” mark of diploma project.