A Novel Maximal Margin Classifier with Application to Logging Lithological Characters Identification

In this paper, by introducing the notion of ``scaled convex hull´´ (SCH) generated by the training points, a novel classifier can be constructed by maximizing the margin between two SCHs when they are separable. Then, fast algorithm to solve the classifier is presented by building the relationship between the SCH and the minimum enclosing ball (MEB). The experiments on the data of logging litho logical characters identification show that the proposed method may achieve better performance than the state-of-the-art methods, in terms of kernel evaluations and execution time.