quaterion.loss.center_loss module¶

class CenterLoss(embedding_size: int, num_groups: int, lambda_c: float | None = 0.5)[source]¶

Bases: GroupLoss

Center Loss as defined in the paper “A Discriminative Feature Learning Approach for Deep Face Recognition” (http://ydwen.github.io/papers/WenECCV16.pdf) It aims to minimize the intra-class variations while keeping the features of different classes separable.

Parameters:

• embedding_size – Output dimension of the encoder.

• num_groups – Number of groups (classes) in the dataset.

• lambda_c – A regularization parameter that controls the contribution of the center loss.

forward(embeddings: Tensor, groups: LongTensor) → Tensor[source]¶

Compute the Center Loss value.

Parameters:

• embeddings – shape (batch_size, vector_length) - Output embeddings from the encoder.

• groups – shape (batch_size,) - Group (class) ids associated with embeddings.

Returns:

Tensor – loss value.

training: bool¶