Fig. 7: 3D bar plot of classifier AUC values (Z-axis) in hold-out validation for each cohort, when considering VaRFS feature sets selected for each combination of regularization parameters for variability (β, X-axis) and sparsity (λ, Y-axis)). | npj Imaging

Fig. 7: 3D bar plot of classifier AUC values (Z-axis) in hold-out validation for each cohort, when considering VaRFS feature sets selected for each combination of regularization parameters for variability (β, X-axis) and sparsity (λ, Y-axis)).

From: Variability Regularized Feature Selection (VaRFS) for optimal identification of robust and discriminable features from medical imaging

Fig. 7

Color shading of the bar plots is based on AUC values such that yellow indicates higher performance while blue corresponds to lower performance. Highest overall AUC performance in each cohort is highlighted in pink, with corresponding regularization parameters summarized.

Back to article page