In this report we describe a novel technique that accelerates learning pirocesses through a dynamic adaptation of the error surface. The algorithm, here name ARON (Adaptive region of Nonlinearity), implements a generalization of the basic McCulloch-Pitts type of neuron which gives to each unit the ability to automatically adapt its operational region according to the requirements of the problem. The changes on the error surface facilitates the progress of the 0ptimization criterion on its search for a minimum. ARON can be used in addition to and bring benefits to a large class of other optimization schemes.
Supervised Learning; Learning Accelerating; Neuron model
Date of this Version