Adaptive recognition by specialized grouping of classes

An algorithm for establishment of class membership conditions, based on making evident the differences among patterns in a labeled training set, is described. Classes in the training set are grouped together in such a way that their exclusive feature values within a group become evident. By making use of these distinctive features and their values, classification of all patterns will be achieved. Identification of faults in a power distribution network is taken as a test case, where after a thorough training, very fast and successful recognition is achieved