Table 2.

The methods and the associated HPs for training the prediction NN. Note that CM-VAE-Adv, CM-VAE-OV, and CM-VAE-Adv-OV are part of the DACMVA framework.

Method	Description	Prediction NN HPs
Multisurv	No augmentation	bs
OV	Oversample from χB with replacement	bs, bsOV, γ
TDImpute	Train EncA, DecB as in Zhou et al. [11]; no oversampling	bs
TDImpute-OV	Train EncA, DecB as in Zhou et al. [11]; with oversampling	bs, bsOV, γ, t
CM-VAE-Adv	With adversarial training; no oversampling (see Section 2.2)	bs
CM-VAE-OV	No adversarial training; with oversampling (see Section 2.3)	bs, bsOV, γ, t
CM-VAE-Adv-OV	With adversarial training; with oversampling (see Section 2.4)	bs, bsOV, γ, t