Comparison of Predictions between Artificial Neural Networks and Gaussian Processes in EMC Investigations

Abstract

Due to the increasing size and complexity of investigations in the field of electromagnetic compatibility, it is a desirable objective to reduce the calculation effort of such studies. In some of them, a large number of similarly constructed DUTs are investigated. For these cases, similar results are expected under equal conditions. Studies have shown, that artificial neural networks (ANNs) can reduce the effort of such investigations. This paper presents a further reduction method, the Gaussian processes. In addition, ANNs and Gaussian processes are compared and the results predictability regarding to EMC are evaluated.