.ce in Networks-A High Order Neural Net Approach
George A. Rovithakis, Athanassios G. Malamos, Theodora A. Varvarigou, Manolis A. Christodoulou · 1998
In this paper, we employ Recurrent High Order Neural Networks (R.HONNs) to determine the unknown values of media characteristics that lead to user satisfaction without violating network limitations. Based on a priori knowledge-measurements, we assume given a nonlinear function that relates media characteristics with user satisfaction, which we further exploit to construct the control error. Based on Lyapunov stability theory weight update laws are developed to guarrantee regulation of the user satisfaction error to zero plus boundedness of all other signals in the closed loop. Simulation studies performed on simple but illustrative examples highlight the approach.