sbi.neural_nets.marginal_nn

Navigation

sbi.neural_nets.marginal_nn#

marginal_nn(model, z_score_x='independent', hidden_features=50, num_transforms=5, num_bins=10, num_components=10, **kwargs)[source]#

Returns a function that builds a density estimator for learning the marginal.

Parameters:
  • model (ZukoFlowType) – The type of density estimator that will be created.

  • z_score_x (Literal['independent', 'structured', 'transform_to_unconstrained', 'none'] | None) – Whether to z-score samples \(x\) before passing them into the network.

  • hidden_features (int) – Number of hidden features.

  • num_transforms (int) – Number of transforms when a flow is used.

  • num_bins (int) – Number of bins used for the splines in nsf.

  • num_components (int) – Number of mixture components for a mixture of Gaussians.

  • **kwargs (Any) – Additional estimator arguments. Valid keys are defined by MarginalFlowConfig; unknown keys trigger a warning and are forwarded to the builder.

Return type:

Callable