Skip to main content
Fig. 2 | Genome Biology

Fig. 2

From: scVAEDer: integrating deep diffusion models and variational autoencoders for single-cell transcriptomics analysis

Fig. 2

scVAEDer accurately learns the latent representation and generates new high-quality scRNA-seq data. a Red, forward diffusion process with 1000 steps on hematopoiesis in zebrafish as we add noise to the data; blue, reverse process as the model learns how to denoise. b UMAP visualization of the real data embedding. c Samples generated from DDM prior. d samples generated from the VAE. e Total variation distance (TVD) between latent embedding of data and samples generated from the DDM as well as VAE prior distributions

Back to article page

Genome Biology

ISSN: 1474-760X

Contact us