Table 2. Hyperparameter combination of the top performing models for each dataset.

Dataset	latent_dim	learning_rate	encoder_batch_ norm	batch_size	val_loss
Cell Painting level 5	10	0.001	True	96	2.24
Cell Painting level 4	90	0.0001	True	32	0.74
L1000	65	0.001	True	512	1363.45