Learning Algorithm and its application to antenna optimization

This paper presents some improvement strategies of Learning Algorithm (LA) including perturbation of the historical best solutions and two hybrid strategies of LA and the Simplex Method (SM). By using 9 benchmark functions of 30-dimensional or/and 2-dimensional domains, numerical experiments are carried out to evaluate the performances of the three improved versions of LA comparing with the original LA, the Genetic Algorithm (GA), the Differential Evolution (DE) and the Particle Swarm Optimization (PSO). Numerical experiment results show that all the three improved versions can achieve obvious performance improvement. Moreover, a patch antenna of nonintuitive planar structures is optimized by LA, which shows that LA has good heuristic search performance and can be applied in antenna optimization.