Fig. 6: Transferring ML model to identifying STZs. | npj Computational Materials

Fig. 6: Transferring ML model to identifying STZs.

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

Fig. 6

a Correlation between the probability estimates, pl, from the L-Eact model and the non-affine displacement (\({\mathrm {D}}^2_{\min}\)) at strain 4.0% of a Cu64Zr36 – 109 K s−1 glass. In the box plots, ends of box spans from 25 to 75% percentile, black line in box represents median, whiskers show 1.5 times the inter-quartile range, and points outside the whiskers show outliers. Outliers are marked in the left plot and are removed in the right plot for clarity. b Receiver operating characteristic (ROC) curve and area under curve (AUC) when using the pl to identify STZs with the largest 5% \({\mathrm{D}}^2_{\min}\).

Back to article page