A Multi-Output Active Learning Method for the Uncertainty Quantification of PCB Lines

This paper introduces a novel active learning method for iteratively building a suitable training set within the scope of Gaussian process models applied to the uncertainty quantification of multi-output system responses. In particular, the proposed strategy offers a significant advantage by extending the already existing generic techniques that typically account for a single scalar output only. This approach extends the training to all outputs, thus improving the accuracy of the overall response of the system in a more comprehensive manner.