Table A1.

Hyperparameters for the WAEs.

	WDN 1	WDN 2	WDN 3
Training hyperparameters
Batch size	256	256	256
Learning rate	$5\times {10}^{-5}$	$5\times {10}^{-5}$	$5\times {10}^{-5}$
$l^2$ regularization	${10}^{-10}$	${10}^{-10}$	${10}^{-10}$
MMD regularization	${10}^{-2}$	${10}^{-2}$	${10}^{-2}$
Scheduler warmup	50	50	50
Early stopping patience	50	50	50
ResNet Architecture
Latent dimension	8	16	16
# layers	5	5	7
# neurons	64, 48, 32,	192, 160, 128,	512, 384, 320,
	24, 16	96, 64, 32	256, 192, 128, 64
Act. func.	Leaky ReLU	Leaky ReLU	Leaky ReLU
Transformer Architecture
Latent dimension	8	16	16
Num. dense neurons	32	128	256
Embedding dimension	4	4	4
# Attn. heads	2	2	2
# Latent transformer blocks	2	2	2
# Full order transformer blocks	1	1	1
Act. func. transformer layers	GeLU	GeLU	GeLU
Act. func. dense layers	Leaky ReLU	Leaky ReLU	Leaky ReLU