LAUSR

In the artificial neural network (ANN), the most time-consuming part for parametric modeling of microwave components is the collection of training datasets from full-wave electromagnetic simulations. However, the reported models for parametric modeling of EM behaviors are based on supervised learning, in which the labeled sampling data from full-wave EM simulations ought to be sufficient for ANN training. Thus, the number of full-wave simulations is the main factor that influences the effectiveness of collecting training and testing samples. Based on the dynamic adjustment kernel extreme learning machine, this paper proposes a semi-supervised learning model lying between supervised learning and unsupervised learning to largely reduce the number of required training samples. The proposed model contains two training processes, the initial training, and the self-training. In the initial training process, a small number of training samples from full-wave simulations are used to make the model rapidly converge. Then, in the self-training process, the model produces unlabeled training datasets to train itself till the testing accuracy is satisfied. Two numerical examples of a microstrip-to-microstrip vertical transition and a dual-band four-pole filter are employed to verify the effectiveness of the semi-supervised learning model.