Max Margin Polyhedral Conic Function Classifier

In classification problems, generalization ability has a key role for successful prediction. Well known Support Vector Machine classifier, tries to increase generalization ability via maximizing the margin, which is the distance between two parallel hyperplanes on the closest points. In this work we investigate maximizing the margin on non-parallel multi surfaces, by adapting GEPSVM* to Polyhedral Conic Function Classifiers.