The training of multilayer perceptron is generally a difficult task. Excessive training times and lack of convergence to an acceptable solution are frequently reported. This paper discribes new training methods of feedforward neural networks. In comparison with standard backpropagation algorithm it ...