Fuzzy Min-Max Neural Networks-Part 1 : Classification

摘要: A supervised learning neural network classifier that utilizes fuzzy sets as pattern classes is described. Each set an aggregate (union) of hyperboxes. hyperbox n-dimensional box defined by a min point and max with corresponding membership function. The min-max points are determined using the algorithm, expansionxontraction process can learn nonlinear class boundaries in single pass through data provides ability to incorporate new refine existing without retraining. use approach classification inherently degree information extremely useful higher level decision mak- ing. This paper will describe relationship between classification. It explains implementation, it outlines recall algorithms, several examples operation demonstrate strong qualities this classifier. AmRN key element many engi- P neering solutions. Sonar, radar, seismic, diagnostic applications all require accurately classify situation. Control, tracking, prediction systems often classifiers determine input-output relationships. Because wide range applicability, has been studied great deal (13), (15), (19). describes creates aggregating smaller into class. technique, introduced (42) extension earlier work (41), data, add on fly, received, uses simple operations allow for quick execution. Fuzzy networks built sets. defines region space patterns full membership. completely its point, function respect these points. (hyperbox) combination set, aggregated form class, resulting structure fits naturally framework; hence system called network. Learning performed properly placing adjusting hyperboxes space.