Physics-inspired machine learning of localized intensive properties

A physics-inspired machine learning approach to predicting localized intensive properties in molecules is presented. The new method is applied to predicting orbital energies and localisations in potential organic semiconductors.