Dimensionality and prototype reduction techniques for pattern analysis

This thesis investigates two important topics in the statistical pattern recognition field, namely dimensionality reduction for supervised classification and prototype reduction for unsupervised classification. For dimensionality reduction part, we concentrate on the Discriminative Linear Dimensionality Reduction (DLDR) techniques with feature extraction for supervised classification as the major application. For prototype reduction part, we focus on the prototype-based clustering algorithms.DOCTOR OF PHILOSOPHY (EEE