Interfacial informatics

Abstract Using machine learning methods to analyse and predict events occurring at interfaces is more complicated than applying participating entities, such as adsorbates adsorbents separately. Whether combining molecular or materials descriptors, explicitly defining the topology of space in between, choice features critical successfully understanding potential energy surface that determines probability events, importance unique characteristics can guide decision making. If reliably defined these descriptors be used advanced model dynamics, co-adsorption interfacial evolution based on complex data; an area traditionally reserved for dynamics kinetic Monte Carlo. In this perspective, we provide some insights into how interactions effectively described introduce relevant go beyond unsupervised pattern recognition supervised classification regression currently preferred by community.