Fig. 1
From: Large-scale evaluation of k-fold cross-validation ensembles for uncertainty estimation
Overview of all predictive performances (\(R^2\)) for the modeling techniques RF (a), XGB (b), SVM (c), SNN (d), and DNN (e) as box-whisker plots. The featurization-specific results (for MACCS, MFC, RDKit descriptors, and CDDD) are shown and colored separately within each plot. The only negative \(R^2\), corresponding to MMP2, using RDKit descriptors and the SNN, was omitted for visual clarification