A probabilistic theory of pattern recognition. (English) Zbl 0853.68150

Applications of Mathematics. 31. New York, NY: Springer. xv, 636 p. (1996).
Non-parametric estimators and bounds for classification rules, \(k\)-nearest neighbor rules, and their consistency; error estimation – Vapnik-Chervonenkis theory about convergence for all data distributions; shatter coefficients and generalized linear discriminants, parametric classification. Tree classifies, BSP trees and splitting criteria. Data dependent partitioning, and resubstitution estimator. Potential kernel functions, automatic kernels and nearest neighbor rules. Neural network classifiers, Hypercube classifiers, Hypercube classifiers, Problems and exercises attached to each chapter.
Reviewer: L.F.Pau (Alvsjo)


68T10 Pattern recognition, speech recognition
68-02 Research exposition (monographs, survey articles) pertaining to computer science