2216 Development of Agent based PSO
Masao ARAKAWA, Satoshi KITAYAMA · Sekkei Kougaku, Shisutemu Bumon Kouenkai kouen rombunshuu/Sekkei Kogaku, Shisutemu Bumon Koenkai koen ronbunshu · 2006
In this paper, we will propose agent based PSO. Basically PSO has three parameters which determine its basic performance. If you give suitable settings, then it moves so nice to satisfy you. However, if you give them ugly, it sometimes converges to local optimum, or it never converge and scatter around global solutions. Many researchers try to adapt these parameters by looking at situation of convergence. However, they still remain in giving all individuals to have same settings. I think these difficulities in setting lie in artificial setting that ignore diversity of individuals. In this study, we allow to give different settings for each individual and allow them to aim different objective function for global search.