Fig. 2: Comparison of metrics calculated from phDOS in the test set.
For each of 3 settings (horizontal shaded bars), the results of the CV (heat capacity at 300 K, left) and \(\bar{\omega }\) (average phonon frequency, right) calculations from the predicted phDOS in the test set are shown as the relative error with respect to the ground truth value for each material in the test set. For each of the 3 settings, 3 models, and 2 properties, the relative errors are shown with a box plot (center line, median; box limits, upper and lower quartiles; whiskers, 1.5× interquartile range; points, outliers). In the MaxNorm-MSE setting, all 3 ML models have similar median relative losses for both CV and \(\bar{\omega }\), with Mat2Spec providing a smaller interquartile range and less extreme outliers. In the other settings, Mat2Spec outperforms the other ML models with respect to median, interquartile range, and outliers.
