torch_concepts.ExogenousVariable

class ExogenousVariable(concepts: List[str], parents: List[Variable | str], distribution: Type[Distribution] | List[Type[Distribution]] | None = None, size: int | List[int] = 1, metadata: Dict[str, Any] | None = None)

Represents an exogenous variable in a concept-based model.

Exogenous variables are high-dimensional representations related to a single endogenous variable. They capture rich, detailed information about a specific concept (e.g., image patches, embeddings, or feature vectors) that can be used to predict or explain the corresponding endogenous concept.

concepts

List of concept names represented by this variable.

Type:

List[str]

parents

List of parent variables in the graphical model.

Type:

List[Variable]

distribution

PyTorch distribution class for this variable.

Type:

Type[Distribution]

size

Dimensionality of the high-dimensional representation.

Type:

int

endogenous_var

The endogenous variable this exogenous variable is related to.

Type:

Optional[EndogenousVariable]

metadata

Additional metadata. Automatically includes ‘variable_type’: ‘exogenous’.

Type:

Dict[str, Any]

Example

>>> from torch.distributions import Normal, Bernoulli
>>> from torch_concepts.distributions import Delta
>>> from torch_concepts import EndogenousVariable, ExogenousVariable
>>> # Endogenous concept
>>> has_wings = EndogenousVariable(
...     concepts='has_wings',
...     parents=[],
...     distribution=Bernoulli,
...     size=1
... )
>>>
>>> # Exogenous high-dim representation for has_wings
>>> wings_features = ExogenousVariable(
...     concepts='wings_exogenous',
...     parents=[],
...     distribution=Delta,
...     size=128,  # 128-dimensional exogenous
... )

__init__(concepts: str | List[str], parents: List[Variable | str], distribution: Type[Distribution] | List[Type[Distribution]] | None = None, size: int | List[int] = 1, endogenous_var: EndogenousVariable | None = None, metadata: Dict[str, Any] | None = None)

Initialize an ExogenousVariable instance.

Parameters:

• concepts – Single concept name or list of concept names.

• parents – List of parent Variable instances.

• distribution – Distribution type (typically Delta or Normal for continuous representations).

• size – Dimensionality of the high-dimensional representation.

• endogenous_var – Optional reference to the related endogenous variable.

• metadata – Optional metadata dictionary.

Methods

__init__(concepts, parents[, distribution, ...])

Initialize an ExogenousVariable instance.

Attributes

in_features	Calculate total input features from all parent variables.
out_features	Calculate the number of output features for this variable.