A semisupervised approach in spherical separation

We embed the concept of spherical separation of two nite disjoint set of points into the semisupervised framework. This approach appears appealing since in the realworld classication problems the number of unlabelled points is very large and labelling data is in general expensive. We come out with a model characterized by an error function which is nonconvex and nondierentiable, that we minimize by means of a bundle method. Numerical results on some small/large datasets drawn from literature are reported.