The elicitation of labeled training data from physical sources is a primary bottleneck that limits the applicability of traditional supervised learning algorithms. Transfer learning algorithms overcome the limitation ...