Including crystal structure attributes in machine learning models of formation energies via Voronoi tessellations

Caption: This figure compares the performance of three machine learning techniques that predict the properties of a material given its crystal structure: the Coulomb Matrix (CM), Partial Radial Distribution Function (PRDF), and the Voronoi-based ML method described in this report. Each model was trained using same set of 30000 entries from the OQMD, and the above plots show the performance of this model on the same validation set of 1000 compounds from the OQMD. Our technique performs better by every metric.