Fig. 7: Plots showing energy error in terms of Hartree/atom across composition space with increasing number of training compositions for SiGeSn (top) and CrFeCoNi (bottom).
From: Electronic structure prediction of medium and high entropy alloys across composition space
Top left: Bulk 64-atom SiGeSn results across composition space, logarithmic scale to emphasize the order of magnitude. Top right: Magnified version of the SiGeSn results, linear scale to emphasize the specific values. Bottom left: Bulk 32-atom CrFeCoNi results across composition space, logarithmic scale. Bottom right: Magnified version of the CrFeCoNi results, linear scale. The dashed lines are present to illustrate the magnification of the magnified plots. Standard deviation bars are shown in each of the plots. The training compositions for Tessellation models are shown in Fig. 3. The training compositions for Active Learning models of SiGeSn are shown in Fig. 4. The training compositions for Active Learning models of CrFeCoNi are shown in Fig. 5.
: Maximum Error (AL), 
: Average Error (AL), 
: Maximum Error (T), 
: Average Error (T).
