A new maximal-margin spherical-structured multi-class support vector machine

摘要: Support vector machines (SVMs), initially proposed for two-class classification problems, have been very successful in pattern recognition problems. For multi-class the standard hyperplane-based SVMs are made by constructing and combining several maximal-margin hyperplanes, each class of data is confined into a certain area constructed those hyperplanes. Instead using hyperspheres that tightly enclosed can be used. Since class-specific separately, spherical-structured used to deal with problem easily. In addition, center radius hypersphere characterize distribution examples from class, may useful dealing imbalance this paper, we incorporate concept maximal margin SVMs. Besides, approach has advantage new parameter on controlling number support vectors. Experimental results show method performs well both artificial benchmark datasets.