TABLE 5.

An overview of all imputation methods and their hyperparameters we optimized. Mean/mode imputation does not have any hyperparameters, and Discriminative DL is optimized using autokeras, which is why we do not explicitly define a hyperparameter grid.

Imputation method	Hyperparameters		Grid size
	Name	Values
Mean/mode	—	—	—
k-NN	$n\_neighbors$	(1, 3, 5)	3
Random forest	$n\_estimators$	(10, 50, 100)	3
Discriminative DL a	—	—	—
VAE	$n\_hidden\_layers$	(0, 1, 2)	3
GAIN	$alpha$	(1, 10)	16
	$hint\_rate$	(0.7, 0.9)	—
	$generator\_learning\_rate$	(0.0001, 0.0005)	—
	$discriminator\_learning\_rate$	(0.00001, 0.00005)	—

^aOptimized using autokeras, see Section 3.4.4.