HYBRID-SURROGATE-MODEL-BASED EFFICIENT GLOBAL OPTIMIZATION FOR HIGH-DIMENSIONAL ANTENNA DESIGN

Efficient global optimization has been extensively used in problems with expensive cost functions. However, this method is not suitable for high-dimensional problems. In this paper, the radial basis function network is introduced into the efficient global optimization, to avoid local optima and achieve a fast convergence for high-dimensional optimization. Our algorithm is applied to a 12-dimensional optimization of a transmitting antenna. Compared to the genetic-algorithm-based efficient global optimization and the differential evolution strategy, our algorithm converges to the global optimal value more efficiently.