Figure 2.
A comparison of estimated surface gravity values from the deep learning regressor, log (g)DL, with asteroseismic surface gravity measurements from the Kepler DR25 stellar properties catalogue (Mathur et al. 2017), log (g)KSPC, for the 1430 stars in our validation sample. The green dashed line indicates a one-to-one relation, while the magenta dashed lines indicate ±0.1 dex deviations from the green line. For clarity, the errorbars of only every fifth blue point is plotted. Marginal detections are plotted in red, while predictions with large deviations from the one-to-one relation (>0.25 dex) that are visually verified to have inaccurately predicted νmax are highlighted in cyan.