A "mutual update" training algorithm for fuzzy adaptive logic control/decision network (FALCON)

The conventional two-stage training algorithm of the fuzzy/neural architect ure called FALCON may not provide accurate results for certain type of prob lems, due to the implicit assumption of independence that this training mak es about parameters of the underlying fuzzy inference system. In this corre spondence, a training scheme is proposed for this fuzzy/neural architecture , which is based on line search methods that have long been used in iterati ve optimization problems. This scheme involves synchronous update of the pa rameters of the architecture corresponding to input and output space partit ions and rules defining the underlying mapping; the magnitude and direction of the update at each iteration is determined using the Armijo rule. In ou r motor fault detection study case, the mutual update algorithm arrived at the steady-state error of the conventional FALCON training algorithm as twi ce as fast and produced a low er steady-state error by an order of magnitud e.