Fuzzy Rule Base Extraction by Bacterial Type Algorithms using selected T-norms

Abstract

In our previous paper we proposed the Trigonometric t-norm and t-conorm and we found that it showed good properties (among various types of tnorms and t-conorms) in Fuzzy Flip-Flop based Neural Networks. In our previous research we have done a number of efforts to improve the efficiency of the bacterial memetic type algorithms in the field of the fuzzy rule base identification. The efficiency of the bacterial type algorithms were tested by Mamdani type inference system. Now, we have examined how the different t-norms used instead of the widely used min operator affect the learning capabilities of the system
applied, the speed of the convergence in case of various training
algorithms. We implemented and tested some promising t-norms
(Algebraic, Hamacher (p=0) and Trigonometric) for the Mamdani type inference system in order to reduce computational effort and processing time. This research is focused on non-parametric t-norms.