Figure 4
From: Kimberlite eruptions driven by slab flux and subduction angle
Cross-validation score and training score (\(R^2\) value) for the nearest-neighbours regression algorithm to predict the dip angle of subducting slabs. The algorithm is tested using increasing numbers of nearest neighbours (k) used in the calculation from Eq. 3. Shaded regions indicate one standard deviation from the mean (solid lines). Cross validation score is shown as negative number for visual clarity to compare the optimal trade-off with the training score.
