抄録
We have been developing multi-point approximation for years for optimization to reduce the number of function calls to obtain adequate optimum results. During these processes, we have been focusing in RBF network, and these days, it has been recognized as one of the effective method as surrogate optimization. Compared with other method such as Krings method, SVM and so on, RBF has the most simple way to form so called response surface, therefore, it has more considerations in parameter settings. In RBF, most significant factor for better approximation lies in setting of radius for each radius function. We have proposed a method for optimization of radius, but it takes too much computation to achieve an adequate results. For that purpose, we have proposed convolution of RBF. However, it needs a lot of parameter settings priori to form response surface. In this study, we are going to propose a new method for compatible Convolute RBF.
