Improvement of Grammatical Swarm
Atsushi UCHIYAMA, Risako Yamamoto, Eisuke Kita · Keisan Rikigaku Koenkai koen ronbunshu/Keisan Rikigaku Kouenkai kouen rombunshuu · 2019
Grammatical Swarm (GS), which is one of the evolutionary computations, is designed to find the function or the program satisfying the design objective. The algorithm of GS has one difficulty. Since the design variables are defined in the real-valued numbers, the real-valued variables have to be rounded to the integer-valued ones. The aim of this study is to present the real-coded Grammatical Swarm (RGS) in order to solve the above problem. The real-coded GS (RGS) is compared with the traditional GS in the symbolic regression problem. The best parameters of both algorithms are determined and then, their results are compared. The results show that RGS can find a better solution than traditional GS.