|View source on GitHub|
Computes the lifted structured loss.
tf.contrib.losses.metric_learning.lifted_struct_loss( labels, embeddings, margin=1.0 )
The loss encourages the positive distances (between a pair of embeddings with the same labels) to be smaller than any negative distances (between a pair of embeddings with different labels) in the mini-batch in a way that is differentiable with respect to the embedding vectors. See: https://arxiv.org/abs/1511.06452.
labels: 1-D tf.int32
Tensorwith shape [batch_size] of multiclass integer labels.
embeddings: 2-D float
Tensorof embedding vectors. Embeddings should not be l2 normalized.
margin: Float, margin term in the loss definition.
lifted_loss: tf.float32 scalar.