Abstract :
It is shown that various classifiers that are based on minimization of a regularized risk are universally consistent, i.e., they can asymptotically learn in every classification task. The role of the loss functions used in these algorithms is considered in detail. As an application of our general framework, several types of support vector machines (SVMs) as well as regularization networks are treated. Our methods combine techniques from stochastics, approximation theory, and functional analysis
Keywords :
approximation theory; functional analysis; learning (artificial intelligence); minimisation; pattern classification; stochastic processes; support vector machines; SVM; approximation theory; asymptotic learning; functional analysis; kernel classifier regularization; minimization; regularization network; stochastic; support vector machine; universal consistency; Approximation methods; Fasteners; Functional analysis; Kernel; Machine learning; Pattern recognition; Statistical distributions; Stochastic processes; Support vector machine classification; Support vector machines; Computational learning theory; kernel methods; pattern recognition; regularization; support vector machines (SVMs); universal consistency;
