cascade.nn

Neural network utilities

Functions

copy_like	Copy tensor device, dtype and data from a source tensor to a target tensor in place
gumbel_sigmoid	Straight-through Gumbel sigmoid sampler
mean_squared_error	Compute the mean squared error along a specified dimension
multi_rbf	RBF kernel with support for multiplex dims
multi_trace	Compute matrix trace with support for multiplex dims

Classes

AcycPrior	Prior that enforces acyclicity constraint
AttnPool	Attention-based pooling layer to combine multiple intervention embeddings
Bilinear	Bilinearly parameterized edge logits
Edgewise	Edgewise parameterized edge logits
EmbLatent	Intervention latent module encoding from fixed embeddings
Func	Structural equation with covariates
GCNLatent	Intervention latent module encoding from a graph
IntervDesign	Intervention design module
Kernel	Abstract class for kernels
KroneckerDelta	Kronecker delta kernel
L1	L1 penalized log prior probability
Latent	Interventional latent module
Likelihood	Abstract class for causal distributions
LogDet	Log-determinant penalized log prior probability
Module	Abstract module class supporting parameter freezing, decayed / non-decayed parameter iteration, and cached property clearing
ModuleList	A module list with the Module capabilities
MultiLinear	Linear layer with support for multi-dims
NegBin	Negative binomial causal distribution
NilLatent	Nil interventional latent module that always outputs the standard normal
Normal	Normal causal distribution
Prior	Compute unnormalized negative log prior probability of a scaffold graph
RBF	Radial basis function kernel
Scaffold	Abstract graph scaffold
ScaleFree	Scale-free penalized log prior probability
SparsePrior	Prior that encourages sparsity
SpecNorm	Spectral norm penalized log prior probability
TrExp	Tr-Exp penalized log prior probability