Fuzzy logic based parameter adjustment model for adaptive decayed brain emotional learning network with application to online time series prediction

Abstract

Adaptive decayed brain emotional learning (ADBEL) is a computationally fast neural network and has shown promising results for online time series prediction problems. However, it is not clear how to set the free parameters namely alpha, beta and gamma of the network. In this work, we present a fuzzy logic based model for adjusting these ADBEL network parameters in an online manner which will not jeopardize the simplicity and quickness of the network. The proposed model uses prediction error and reward signals as inputs and produces the network parameters as outputs. Further, the fuzzy logic based model employs a small rule base to vary the network parameters in order to maintain the complexity of the ADBEL network at minimum. Few chaotic time series namely the Mackey glass, Lorenz, Rossler, and the disturbance storm time index are used to check the validity of the fuzzy logic based parameter adjustment model for ADBEL network using MATLAB® programming environment.