Interaction

class Interaction(*args, **kwargs)[source]

Bases: Module, Generic[HeadRepresentation, RelationRepresentation, TailRepresentation], ABC

Base class for interaction functions.

Initialize internal Module state, shared by both nn.Module and ScriptModule.

Attributes Summary

`dimensions`	Get all the relevant dimension keys.
`entity_shape`	The symbolic shapes for entity representations
`head_indices`	Return the entity representation indices used for the head representations.
`head_shape`	Return the symbolic shape for head entity representations.
`is_complex`	whether the interaction is defined on complex input
`relation_shape`	The symbolic shapes for relation representations
`tail_indices`	Return the entity representation indices used for the tail representations.
`tail_shape`	Return the symbolic shape for tail entity representations.
`value_range`	the interaction's value range (for unrestricted input)

Methods Summary

`forward`(h, r, t)	Compute broadcasted triple scores given broadcasted representations for head, relation and tails.
`reset_parameters`()	Reset parameters the interaction function may have.
`score`(h, r, t[, slice_size, slice_dim])	Compute broadcasted triple scores with optional slicing.
`score_h`(all_entities, r, t[, slice_size])	Score all head entities.
`score_hrt`(h, r, t)	Score a batch of triples.
`score_r`(h, all_relations, t[, slice_size])	Score all relations.
`score_t`(h, r, all_entities[, slice_size])	Score all tail entities.

Attributes Documentation

dimensions

Get all the relevant dimension keys.

This draws from Interaction.entity_shape, and Interaction.relation_shape.

Returns:: a set of strings representing the dimension keys.

entity_shape: Sequence[str] = ('d',): The symbolic shapes for entity representations

head_indices: Return the entity representation indices used for the head representations.

head_shape: Return the symbolic shape for head entity representations.

is_complex: ClassVar[bool] = False: whether the interaction is defined on complex input

relation_shape: Sequence[str] = ('d',): The symbolic shapes for relation representations

tail_indices: Return the entity representation indices used for the tail representations.

tail_shape: Return the symbolic shape for tail entity representations.

value_range: ClassVar[ValueRange] = ValueRange(lower=None, lower_inclusive=False, upper=None, upper_inclusive=False): the interaction’s value range (for unrestricted input)

Methods Documentation

abstract forward(h: HeadRepresentation, r: RelationRepresentation, t: TailRepresentation) → Tensor[source]

Compute broadcasted triple scores given broadcasted representations for head, relation and tails.

In general, each interaction function (class) expects a certain format for each of head, relation and tail representations. This format is composed of the number and the shape of the representations.

Many simple interaction functions such as TransEInteraction operate on a single representation, however there are also interactions such as TransDInteraction, which requires two representations for each slot, or PairREInteraction, which requires two relation representations, but only a single representation for head and tail entity respectively.

Each individual representation has a shape. This can be a simple \(d\)-dimensional vector, but also comprise matrices, or even high-order tensors.

This method supports the general batched calculation, i.e., each of the representations can have a preceding batch dimensions. Those batch dimensions do not necessarily need to be exactly the same, but they need to be broadcastable. A good explanation of broadcasting rules can be found in NumPy’s documentation.