Support vector machine and its difficulties from control field of view

The application of the support vector machine (SVM) classification algorithm to large-scale datasets is limited due its use a large number vectors and dependency performance on kernel parameter. In this paper, SVM redefined as control system iterative learning (ILC) method used optimize SVM’s ILC technique first defines an error equation then iteratively updates function regularization parameter using training previous state system. closed loop structure proposed increases robustness uncertainty improves convergence speed. Experimental results were generated nine standard benchmark covering wide range applications. show that generates superior or very competitive in term accuracy than those classical state-of-the-art based techniques while significantly smaller vectors.