Fig. 3: Performance of the Ens-3 and Ens-6 models on the test set.
From: Outlier-detection for reactive machine learned potential energy surfaces
A, B on the left show residual plots of the error between reference and prediction. The 1000 energies with the largest variance are shaded with a different color and directly reflect the model’s capability to detect outliers. The corresponding color bar (blue to green/red to yellow) represents the scale of the variance. Note that the variance of most of the test samples is at the lower end of the color bar. Squared error distribution (solid lines) and variance distributions (dotted lines) are shown in the centre next to A, B for comparison. Complementary to this is the variance distribution shown on the right of both panels. Notice that the x axis on the centre and right are in logarithmic scale.
