scglue.models.sc.NBDataDecoder.forward

NBDataDecoder.forward(u, v, b, l)

Decode data from sample and feature latent

Parameters:

• u (torch.Tensor) – Sample latent

• v (torch.Tensor) – Feature latent

• b (torch.Tensor) – Batch index

• l (torch.Tensor) – Optional normalizer

Return type:

torch.distributions.negative_binomial.NegativeBinomial

Returns:

recon – Data reconstruction distribution