A novel SVM Geometric Algorithm based on Reduced Convex Hulls

Geometric methods are very intuitive and provide a theoretically solid viewpoint to many optimization problems. SVM is a typical optimization task that has attracted a lot of attention over the recent years in many pattern recognition and machine learning tasks. In this work, we exploit recent results in reduced convex hulls (RCH) and apply them to a nearest point algorithm (NPA) leading to an elegant and efficient solution to the general (linear and nonlinear, separable and non-separable) SVM classification task