Applications of self-organization networks spatially isomorphic to patterns

A new technique based on self-organization is proposed for classifying patt erns (which include characters, and two- and three-dimensional objects). A neuronal network, created to be a physical replica of each exemplar, is map ped onto the given test pattern by self-organization, during which the netw ork undergoes deformation in an attempt to match the given test pattern. Th e extent of deformation is inversely proportional to the correctness of the match: smaller the deformation, better is the match. A deformation measure is proposed, leading to the classification of the test pattern. Also prese nted are some algorithmic improvements (including the choice of other defor mation measures) to speed up computation. Examples illustrate the versatili ty of the technique. (C) 1999 Elsevier Science Inc. All rights reserved.