Support vector machines: heuristic of alternatives
Marcin Orchel · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2007
In this paper it will be presented Sequential Minimal Optimization (SMO) default heuristic optimization. SMO is an algorithm for solving Support Vector Machines (SVM) problem. SMO default heuristic chooses to the active set the worst two parameters based on the Karush-Kuhn-Tucker (KKT) conditions. The proposed heuristic of alternatives chooses parameters to the active set on the basis of not only KKT conditions, but also objective function value growth. Tests show that heuristic of alternatives is generally better than SMO default heuristic.