nichecompass.modules.compute_gp_l1_reg_loss
- nichecompass.modules.compute_gp_l1_reg_loss(model, gp_type, l1_targets_mask=None, l1_sources_mask=None)
Compute L1 regularization loss for the rna decoder weights of gene programs of the type ´gp_type´ to encourage gene sparsity of those gene programs.
- Parameters:
model (
Module) – The VGPGAE module.gp_type (
Literal['prior','addon']) – Type of gene programs to which the L1 regularization loss should be applied.l1_targets_mask (
Optional[Tensor] (default:None)) – Boolean gene program gene mask that is True for all gene program target genes to which the L1 regularization loss should be applied (dim: n_genes, n_gps).l1_sources_mask (
Optional[Tensor] (default:None)) – Boolean gene program gene mask that is True for all gene program source genes to which the L1 regularization loss should be applied (dim: n_genes, n_gps).
- Return type:
Tensor- Returns:
gp_l1_reg_loss: L1 regularization loss for the rna decoder weights.