Adaptive inverse disturbance canceling control system based on least square support vector machines

Adaptive inverse disturbance canceling control uses some adaptive filters. The neural network methods of training these filters have been fully researched. However, the problems of local minimum, curse of dimensionality and overfitting limit the application of neural networks. Comparatively, Support Vector Machines effectively overcome these limitations. A kind of adaptive inverse disturbance canceling control system based on least squares support vector machines (LS-SVM) is proposed. The approach of modeling and inverse modeling using LS-SVM is presented. A parameter selecting method within the Bayesian evidence framework is given for SVM regression with Gaussian kernel. Simulation results show that the approach is effective.