Input-Sensitive Fuzzy Cognitive Maps

Complex systems with nonlinearities and surrounding uncertainty are usually modeled sufficiently by Fuzzy Cognitive Maps (FCMs). FCMs work efficiently even with missing data. Experts, for each case study, support with their knowledge the developed FCMs. Nevertheless, the main drawback of FCMs is their convergence to the same equilibrium point regardless of the initial conditions. In this paper a different approach for modeling FCMs is proposed, where the inputs gain back their lost importance. Thus, Input-Sensitive Fuzzy Cognitive Maps (IS-FCMs), supported both by experts and by the appropriate Rule-Base, manage to converge to desired operating points. The Nonlinear Hebbian Learning algorithm (NHL) is used in order to optimize the values of the weights. A PV-System application is presented. The simulation results support the hypothesis of the proposed new IS-FCM model.