Artificial Neural Network (ANN) have been recently proposed as a mean to speed up the optimized design procedure of printed Reflectarrays, creating a surrogate model of a patch radiator as a function of its geometric parameters, the angle of incidence and frequency. This paper presents an improvement of ANN learning procedure by hybridising classical Error Back-Propagation with Meta Particle Swarm Optimization algorithm. In this way the ANN learning procedure proved to converge in a much more effective way, i.e. with the necessity of the introduction of a smaller size set of training samples and with a significant reduction of the computational effort and of the data memory storage.
Modeling of reflectarray elements by means of MetaPSO-based Artificial Neural Network
MUSSETTA, MARCO;ZICH, RICCARDO
2013-01-01
Abstract
Artificial Neural Network (ANN) have been recently proposed as a mean to speed up the optimized design procedure of printed Reflectarrays, creating a surrogate model of a patch radiator as a function of its geometric parameters, the angle of incidence and frequency. This paper presents an improvement of ANN learning procedure by hybridising classical Error Back-Propagation with Meta Particle Swarm Optimization algorithm. In this way the ANN learning procedure proved to converge in a much more effective way, i.e. with the necessity of the introduction of a smaller size set of training samples and with a significant reduction of the computational effort and of the data memory storage.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.