关键词: 支持向量机, 逆控制, 有限增益稳定

Abstract: Support Vector Machines (SVM) is a new machine learning method on Structural Risk Minimization (SRM) with good generability. The method of constructing inverse controller using SVM was given in this paper. A pseudo-linear compound system was formed when cascading with the inverse controller before the original system. The finite-gain stability of the controller was proved under the assumption that the kernel function was local Lipschitz in this open-loop inverse control system. The sufficient condition for that Gaussian kernel function to the local Lipschitz of either variable was given. The stability of the whole system was also proved under some proper assumptions.

Key words: Support Vector Machines (SVM), inverse control, finite-gain stability