Fig. 2: Predicting the heterogeneity of thermally activated events. | npj Computational Materials

Fig. 2: Predicting the heterogeneity of thermally activated events.

From: Predicting the propensity for thermally activated β events in metallic glasses via interpretable machine learning

Fig. 2

a Receiver operating characteristic (ROC) curve and area under curve (AUC) in classifying the atoms showing the highest 5% (H-Eact problem) or lowest 5% activation energy (L-Eact problem). The dashed line marks a random case. b Near-perfect calibration of the ML-evaluated class probability estimates, that is, ph from H-Eact and pl from L-Eact.

Back to article page