Neural network training schemes for antenna optimization

Thanks to the advantage of low profile and low cost, microstrip ring antenna design has been an interesting and challenging issue in modern engineering society. The trade-off among all the degrees of freedom becomes quite complex and direct antenna synthesis by full-wave analysis are often not applicable. In optimization scheme, the associated cost function by computational approach is always expensive and time-consuming. Artificial Neural Network (ANN) has been exploit as a modeling methodology in Electromagnetic field in recent years. In this article, a new approach with the aim of boosting “online-trading information” between the global optimizer and ANN surrogate model will be discussed.