Learning To Classify

Abstract

The objective of the following chapter is to communicate our experience about the attempt to learn to classify. First we will present a general framework for the classification problem. The classical pattern recognition model is extended in the sense that new ideas about a more global understanding of the classification problem are outlined. Then we will present the methods and tools that were developed in the context of the work that was being done, namely the pattern recognition toolbox TOOLDIAG and the supervised learning algorithm Q*. Finally we will apply the results of the previous theories in the context of Machine tool supervision.