Support Vector Machines

There is a growing interest in support vector machines (SVMs) as tools in ecological prediction and classification. SVMs use separating hyperplanes and exploit theorems bounding the actual risk by the empirical risk to find the optimal decision rule. SVMs do not rely on any distributional assumptions or asymptotic arguments and therefore can be used on small data sets. They are especially valuable for high-dimensional data. This article introduces the underlying mathematical theory of SVMs and discusses their use in ecological investigations.