Table 1.

Mean absolute error (standard error) and Spearman correlation coefficient (ρ) between age and prediction error for each model on the validation set.

		BLUP-mean	BLUP-quantiles	SVM	6-layer CNN	Age spe. 6-layer CNN	ResNet	Inception V1	Ensemble prediction	PAC results
First challenge	MAE (SE)	5.32 (0.19)	4.90 (0.19)	5.31 (0.18)	4.18 (0.16)	4.01 (0.15)	4.02 (0.15)	3.82 (0.14)	3.46 (0.13)*	3.33
	|ρ|	0.32	0.37	0.58	0.25	0.30	0.24	0.41	0.32	0.21
Second challenge	MAE (SE)	6.15 (0.23)	5.96 (0.23)	6.14 (0.23)	5.27 (0.21)	5.17 (0.20)	5.25 (0.20)	4.97 (0.19)	4.69 (0.19)*	4.83
	|ρ|	0.14	0.15	0.15	0.084	0.068	0.11	0.058	0.058	0.021

The standard error [SE = SD/sqrt(N)] reflects the uncertainty around the MAE estimate. A 95% confidence interval may be calculated as MAE ± 1.96*SE, though it (falsely) assumes normality of the absolute error distribution. We performed ensemble prediction using linear combination of age predictors, with linear weights estimated via linear regression. SE of the MAE for ensemble prediction were calculated by bootstrap.

^*Indicates a significant reduction of MAE via ensemble learning compared with Inception alone (p <0.05). PAC results were provided by the PAC team and estimated on participants not available to the authors.