Fig. 11
From: Large-scale evaluation of k-fold cross-validation ensembles for uncertainty estimation
Comparisons between single DNNs and DNN ensembles of 200 members obtained by 2-fold CV. Single DNN performances are colored in blue, performances by DNN ensembles in orange. The upper row shows the comparisons between predictive performances (\(R^2\)) for MACCS (a), MFC (b), RDKit descriptors (c), and CDDD (d). The lower row shows the comparisons between UQ performances (\(\rho\)) for MACCS (e), MFC (f), RDKit descriptors (g), and CDDD (h)