Fig. 8: Learning the efficiency of Cartesian atomic cluster expansion (CACE) long-range (CACE-LR) models trained only on bulk electrolyte solution or electrolyte-vapor interfacial configurations of potassium fluoride (KF) aqueous solution.
From: Machine learning of charges and long-range interactions from energies and forces

The mean absolute errors (MAEs) on forces (F) and charges (q) are shown for CACE-LR models. Both panels display the model performance as a function of training set size (N) for each configuration type.