Linguistic information feed-forward-based dynamical fuzzy systems. I. Theory

This work consists of two parts: theory and evaluation. In the first part, we propose a linguistic information feedforward-based dynamical fuzzy system (LIFFDFS), in which the past fuzzy inference output in terms of membership function is fed forward locally with trainable feedforward parameters. The LIFFDFS can overcome the common static mapping drawback of conventional fuzzy systems. We give a detailed description of its principle and structure. Based on the gradient descent method, an adaptive learning algorithm for the feedforward parameter is also derived. Part 2 of this work discusses applying the proposed LIFFDFS in time series prediction. The well-known Box-Jenkins gas furnace data is used as an evaluation example. Simulation results demonstrate that our fuzzy model has the advantage of inherent dynamics, and is therefore well suited for handling temporal problems like process modeling and control.