Training Trajectories by Continuous
Lutz Leistritz, Miroslaw Galicki, Herbert Witte, Eberhard Friedrich Kochs · 2002
This paper addresses the problem of training trajec- tories by means of continuous recurrent neural networks whose feedforward parts are multilayer perceptrons. Such networks can approximate a general nonlinear dynamic system with arbitrary accuracy. The learning process is transformed into an optimal con- trol framework where the weights are the controls to be deter- mined. A training algorithm based upon a variational formulation of Pontryagin's maximum principle is proposed for such networks. Computer examples demonstrating the efficiency of the given ap- proach are also presented.