Very often, projects entail the investigation of several similar systems, e.g. the same molecule on Cu, Ag, and Au, or molecules with similar shape and interaction. Presently, our approach requires to apply machine learning for each system from scratch. However, clearly, there is something that can be learned from one system that can be transferred to the next. The aim of this topic is to combine physical insight with the description of the structures within our machine learning model, in order to transfer knowledge between different systems.