Nonparametric classification using radial basis function nets andempirical risk minimization

摘要

We study convergence properties of radial basis function (RBF)networks in nonparametric classification for a large class of basisfunctions with parameters of RBF nets learned through empirical riskminimization. In the classification (pattern recognition) problem, basedupon the observation of a random vector X∈Rd,one has to guess the value of a corresponding label Y, where Y is arandom variable taking its values from (-1,1)