besskge.loss.MarginBasedLossFunction

class besskge.loss.MarginBasedLossFunction(margin, negative_adversarial_sampling, negative_adversarial_scale=1.0, loss_scale=1.0)[source]

Base class for margin-based loss functions.

Initialize margin-based loss function.

Parameters:

margin (float) – The margin to be used in the loss computation.
negative_adversarial_sampling (bool) – see BaseLossFunction
negative_adversarial_scale (float) – see BaseLossFunction
loss_scale (float) – see BaseLossFunction

abstract forward(positive_score, negative_score, triple_weight)

Compute batch loss.

Parameters:

positive_score (Tensor) – shape: (batch_size,) Scores of positive triples.
negative_score (Tensor) – shape: (batch_size, n_negative) Scores of negative triples.
triple_weight (Tensor) – shape: (batch_size,) or () Weights of positive triples.

Return type:

Tensor

Returns:

The batch loss.

get_negative_weights(negative_score)

Construct weights of negative samples, based on their score.

Parameters:: negative_score (Tensor) – : (batch_size, n_negative) Scores of negative samples.
Return type:: Tensor
Returns:: shape: (batch_size, n_negative) if BaseLossFunction.negative_adversarial_sampling else () Weights of negative samples.

loss_scale: Tensor: Loss scaling factor, might be needed when using FP16 weights

negative_adversarial_sampling: bool: Use self-adversarial weighting of negative samples.

negative_adversarial_scale: Tensor: Reciprocal temperature of self-adversarial weighting