Learning on Scarce Data for Industrial Control: a Transfer Learning approach

Learning on Scarce Data has become a popular topic of research during the last years. Machine Learning (ML) algorithms have shown excellent performance at different application domains but the need of big training datasets is mandatory in most cases. For this reason, an important area of research is focusing on how exploiting ML benefits in scenarios with more limited (scarce) data. This work has a goal to shed some light on this topic when addressing data-based controllers. We give emphasis on Transfer Learning as an efficient tool by providing an application example based on a Waste-Water Treatment Plant (WWTP) scenario.