A self-organizing supervised classifier

A new supervised neural network classifier for online learning is introduced. An association of prototype neurons and fuzzy membership function (MF) is used for cluster approximation. The new architecture based on adaptive resonance theory (ART) dedicates one adapted ART module (ARTMOD) to each class of patterns. Each prototype neuron defines a hyper-sphere in the input space. A class consists of a collection of hyper-spheres. Learning is stable since the hyper-spheres can only expand until they encompass a pattern of another class, then their expansion is definitely stopped. Each prototype neuron is associated with a vigilance threshold that can be set to arbitrary low initial value and still ensure minimum error rate learning. Misclassification increases the threshold of the active neuron by the minimum amount to correct a predictive error. A new fuzzy membership function is introduced. Its kernel shape allows an adaptive confidence rate variation while according a constant rate to every training pattern.