A flexible classification approach with optimal generalisation performance: support vector machines

Support vector machines (SVM) as a recent approach to classification implement classifiers of an adjustable flexibility, which are automatically and in a principled way optimised on the training data for a good generalisation performance.

Abstract

Measuring a larger number of variables simultaneously becomes more and more easy and thus widespread. Obtaining a sufficient number of training samples or measurements, on the other hand, is still time-consuming and costly in many cases. Therefore, the problem of efficient learning from a limited training set becomes increasingly important. Support vector machines (SVM) as a recent approach to classification address this issue within the framework of statistical learning theory. They implement classifiers of an adjustable flexibility, which is automatically and in a principled way, optimised on the training data for a good generalisation performance. The approach is introduced and its learning behaviour examined.