Table 3.

Mean absolute error (in years) and standard deviation of the models trained during cross-validation. The best absolute value for each modeling strategy is marked in bold

Modeling strategy	Learning rate
	9*10−4	6*10−4	3*10−4	1*10−4	9*10−5	6*10−5
Single best model	1.01 ± 0.84	0.98 ± 0.8	0.98 ± 0.8	0.96 ± 0.8	0.95 ± 0.79	0.97 ± 0.81
Snapshot ensembling	0.97 ± 0.81	0.94 ± 0.79	0.94 ± 0.78	0.92 ± 0.76	0.93 ± 0.78	0.95 ± 0.78