Support Vector Array Processing

Support vector machines (SVM) have improved generalization performance over other classical optimization techniques due to the Thikonov regularization scheme included in its formulation. In this paper, an SVM-based approach for nonlinear array processing which is inspired on the minimum variance distortionless method (MVDM) is introduced. The approach is developed through the formulation of the MVDM plus the SVM regularization in a kernel reproducing Hilbert space, which leads to a nonlinear counterpart of the MVDM. Comparison examples are included to show the validity of the new minimization approach