Classification in the multiple instance learning framework via spherical separation

We consider a multiple instance learning problem where the objective is the binary classifications of bags of instances, instead of single ones. We adopt spherical separation as a classification tool and come out with an optimization model which is of difference-of-convex type. We tackle the model by resorting to a specialized nonsmooth optimization algorithm, recently proposed in the literature which is based on objective function linearization and bundling. The results obtained by applying the proposed approach to some benchmark test problems are also reported.