University of Leuven, Belgium
“Statistical Learning Theory”
Statistical learning is engaged in the discovery of regularities from empirical data. In the case of supervised learning these data have, for example, an input component X and an output component Y, and the goal of learning is then to describe the relationship of Y and X. It is essential here that this relationship is true not only for the past but also for future observations. For unsupervised learning, the output component is not required, so that the general properties of the distribution P of the data should be found out. Since non-parametric learning algorithms achieve this goal without preliminary assumptions about the shape of the distribution P, statistical learning methods play an increasingly important role in the solution of complex problems. This is reflected in many applications in, for example, bioinformatics, medicine, the economy (Data Mining), the engineering sciences, and, last but not least, Internet search engines.