Statistical Pattern Recognition

Andrew Webb's graduate-level text on statistical pattern recognition, covering density estimation, linear and non-linear discriminant analysis, dimensionality reduction, clustering, and the evaluation of classifiers. The text set out the statistical foundations on which much of modern machine-learning practice is built.

Chapters move from Bayes-rule classification and parametric and non-parametric density estimation, through linear and non-linear discriminant functions, neural network classifiers, support-vector machines and tree-based methods, into dimensionality reduction by principal-component and projection methods, feature selection, clustering and the assessment of classifier error rate. The treatment emphasises feature selection, error estimation and the curse of dimensionality.

Andrew Webb spent his professional career at the UK Defence Science & Technology Laboratory, where high-dimensional sensor data from electro-optic and radar systems demanded exactly the statistical machinery later popularised under the machine-learning label. Successive editions have been carried forward under Wiley after the Hodder Arnold transfer.