Table 4.

The MAE for the models trained with different loss functions.

Age range	Sample size, $N$	Absolute loss training [MAE, (years)]	Relative loss training [MAE, (years)]
[5,20]	281	2.45	2.39
[20,46]	528	5.78	5.54
[46,88]	526	8.01	8.34
Full dataset	1,335	5.96	5.98

The table shows the found age groups and their size, with the corresponding MAE (in years; lower is better) metrics of two models. Square brackets indicate that the end of the range is inclusive; parentheses indicate that the end of the range is exclusive.