Fig. 4: Effect of individual DA selector conditions for SOAP-based model.
From: Identifying domains of applicability of machine learning models for materials science
The distributions of the three features referenced in the selector σSOAP ≡ a/c ≤ 3.87 ∧ γ < 90.35∘ ∧ β ≥ 88.68∘, i.e., a/c (a), γ (b), and β (c), are shown with subpopulation selected by condition in blue, subpopulation deselected in gray, and threshold by red line.
