Table 2.

Zebrafish.

Clustering method	Dim size (m)	Latent representation				T-SNE embedding of latent representation
		DiffVAE	VAE	AE	PCA	DiffVAE	VAE	AE	PCA
k-means	20	0.803	0.771	0.799	0.633	0.809	0.738	0.699	0.717
	50	0.829	0.775	0.811	0.629	0.831	0.801	0.759	0.709
	100	0.844	0.831	0.815	0.627	0.815	0.796	0.806	0.680
DBSCAN	20	0.007	0.004	0.001	0.0002	0.753	0.717	0.556	0.506
	50	0.474	0.243	0.223	0.0009	0.710	0.667	0.573	0.590
	100	0.154	0.018	0.011	0.002	0.813	0.799	0.749	0.570

Mean ARI obtained for clustering the latent representation and the t-SNE embedding of the latent representation for three settings of the reduced dimension size m. The clustering algorithms used are k-means and Gaussian Mixture Models.