ConvKBInteraction

class ConvKBInteraction(hidden_dropout_rate=0.0, embedding_dim=200, num_filters=400)

Bases: FunctionalInteraction[FloatTensor, FloatTensor, FloatTensor]

A stateful module for the ConvKB interaction function.

See also

pykeen.nn.functional.convkb_interaction`()

Initialize the interaction module.

Parameters:

• hidden_dropout_rate (float) – the dropout rate applied on the hidden layer

• embedding_dim (int) – the entity and relation embedding dimension

• num_filters (int) – the number of filters (=output channels) of the convolution

Methods Summary

func(r, t, conv, activation, hidden_dropout, ...)	Evaluate the ConvKB interaction function.
reset_parameters()	Reset parameters the interaction function may have.

Methods Documentation

func(r, t, conv, activation, hidden_dropout, linear)

Evaluate the ConvKB interaction function.

\[W_L drop(act(W_C \ast ([h; r; t]) + b_C)) + b_L\]

Parameters:

• h (FloatTensor) – shape: (*batch_dims, dim) The head representations.

• r (FloatTensor) – shape: (*batch_dims, dim) The relation representations.

• t (FloatTensor) – shape: (*batch_dims, dim) The tail representations.

• conv (Conv2d) – The 3x1 convolution.

• activation (Module) – The activation function.

• hidden_dropout (Dropout) – The dropout layer applied to the hidden activations.

• linear (Linear) – The final linear layer.

Return type:

FloatTensor

Returns:

shape: batch_dims The scores.

reset_parameters()

Reset parameters the interaction function may have.