A neural network approach to adaptive pattern analysis - the deformable feature map

In this paper, we presen t an algorithm that provides adaptive plasticit y in function approximation problems: the deformable (feature) map (DM) algorithm. The DM approach reduces a class of similar function approximation problems to the explicit supervised one-shot training of a single data set. This is followed by asubsequen t, appropriate similarity transformation whic his based on a self-organized deformation of the underlying multidimensional probability distributions. After discussing the theory of the DM algorithm, w e use a computer sim ulation to visualize its e ects on a t w o-dimensional toy example. Finally, we presen t results of its application to the real-world problem of fully automatic voxel-based multispectral image segmentation, employing magnetic resonance data sets of the human brain.