Fig. 4. Box plot representation of the MLPs' learning performance across the compositional space of the MD17 database. All ML models were trained only on energies. Each box plot represents the distribution of the root-mean-squared error (RMSE) for the total energies of all molecules in the MD17 database calculated with respect to the true labels of a test set with 20k geometries. The white dots in the box plots correspond to the mean value of the RMSE for all different molecules.