Advanced Bacterial Memetic Algorithms

Abstract

Fuzzy systems have been successfully used in the arca of controllers for a
long time. The fiist appearance of these controllers was in 1974 by Mamdani and Assilian. The main problem intheusage of Mamdoni-type iriference system and other fuzzy logic based controllers is how to gain the fuzzy rule base (FRB) what the inference system based on. Nawa, Hashiyama, Fumhashi and Uchikawa proposed a novel type of Genetic Algorithm cailed Pseudo-Bacterial Genetic Algorithm (PBGA) for fuzzy rule base (FRB) extraction from input-output data (1995, 1997). Furthermore, Nawa and Furuhashi improved the PBGA caneerning the relationship among the individuals (the rules) and proposed a new method for automatic FRB identification named Bacterial Evolutionary Algorithm
(BEA) (1999). Botzheim, Cabrita, Kóczy and Ruano have applied another technique to gain FRBs, they have used a local search method, the Levenberg-Marquardt algorithm (LM). The most significant improvement in the speed of convergence and the quality of the model achieved by the FRB identification process was the idea of comhining the global and local search methods. Botzheim, Cabrita, Kóczy and Ruano proposed a novel method called Bacterial Memetíc Algorithm (BMA, 2005). Although Bacterial Memetic Algorithm provicles a very good speed of convergence towards the optimal rule base there are likely some po ints of the algorithm where the performanec could be increased. Gál, Botzheim, Kóczy and Ruano proposed new elements (Swap, Merge) for knot order vialation handling in Levenberg-Marquardt method used in BMA (Improved Bacterial Memetic Algorithm, IBMA, 2008) [6]. Another improvement of the BMA is the Bacterial Memetic Algorithm with the modified operation execution order
(BMAM, Gál, Botzheim and Kóczy, 2008). This new approach proposed
using the Levenberg-Marquardt method more efficiently. Comhining the
improvements in IBMA and BMAM is beneficial (Gál, Botzheim and Kóczy,
2008). This advanced version of the Bacterial Memetic Algorithm used
fór FRB extraction named Modified Bacterial Memetic Algorithm. This paper summarízes the bacterial type evolutionary algorithms used for FRB identification.