Principles of nonparametric learning. Papers from the summer school, Udine, Italy, July 9--13, 2001 (Q1407941)

The goal of this book is to describe a new approach to dependency estimation problems which originated within nonparametric learning. This theory is used in various fields of applications, like pattern recognition, density and regression function estimation, data compression, on-line learning, prediction, etc. The book consists of six independent, but closely related areas of nonparametric inference with emphasis on the cited areas. The first chapter begins by considering the basics of statistical learning theory and pattern recognition, including: Vapnik-Chervonenkis theory, complexity regularization, error estimation, etc. Chapter 2 describes nonparametric regression function estimation, such as local averaging estimates, empirical risk minimization applied to spline estimates, and penalized least squares estimates. Chapter 3 gives the theory of on-line prediction of individual sequences. Chapter 4 describes nonparametric learning in vector quantization and clustering, both for fixed-rate and variable-rate data compression. In chapter 5, the study of the \(L_1\)-theory of density estimation is summarized. Chapter 6 reviews nonparametric learning based on evolution algorithms and genetic algorithms with grammars. This book is a professional reference for researchers and practitioners in statistics. It is also intended as classroom resource for graduate courses in statistics and in computer sciences.