StructuredEmbeddingInteraction

class StructuredEmbeddingInteraction(p, power_norm=False)[source]

Bases: pykeen.nn.modules.TranslationalInteraction[torch.FloatTensor, Tuple[torch.FloatTensor, torch.FloatTensor], torch.FloatTensor]

A stateful module for the Structured Embedding (SE) interaction function.

See also

pykeen.nn.functional.structured_embedding_interaction()

Initialize the translational interaction function.

Parameters

  • p (int) – The norm used with torch.norm(). Typically is 1 or 2.

  • power_norm (bool) – Whether to use the p-th power of the L_p norm. It has the advantage of being differentiable around 0, and numerically more stable.

Attributes Summary

relation_shape

The symbolic shapes for relation representations

Methods Summary

func(r_h, r_t, t, p[, power_norm])

Evaluate the Structured Embedding interaction function.

Attributes Documentation

relation_shape: Sequence[str] = ('dd', 'dd')

The symbolic shapes for relation representations

Methods Documentation

func(r_h, r_t, t, p, power_norm=False)

Evaluate the Structured Embedding interaction function.

\[f(h, r, t) = -\|R_h h - R_t t\|\]
Parameters

  • h (FloatTensor) – shape: (batch_size, num_heads, 1, 1, dim) The head representations.

  • r_h (FloatTensor) – shape: (batch_size, 1, num_relations, 1, rel_dim, dim) The relation-specific head projection.

  • r_t (FloatTensor) – shape: (batch_size, 1, num_relations, 1, rel_dim, dim) The relation-specific tail projection.

  • t (FloatTensor) – shape: (batch_size, 1, 1, num_tails, dim) The tail representations.

  • p (int) – The p for the norm. cf. torch.norm.

  • power_norm (bool) – Whether to return the powered norm.

Return type

FloatTensor

Returns

shape: (batch_size, num_heads, num_relations, num_tails) The scores.