A Review of Statistical Learning Theory

Abstract

Statistical learning theory is developed on the basis of machine learning, and it provides the theoretical basis under the condition of small samples of pattern recognition, function fitting and the density estimation. In recent years, the research on statistical learning theory has made more and more important contribution to the research on artificial intelligence. In view of its importance, this paper has sorted out the basic problems of statistical learning theory :(1) The conditions of statistical learning consistency under the rule of minimum empirical risk minimization;(2)The conclusions about the generalization of statistical learning methods in these conditions;(3) Small sample induction and inference criteria established on the basis of these boundaries;(4) At last, the current research status and the further development of statistical learning theory are discussed.