Computes the lifted structured loss.
loss_lifted_struct(margin = 1, name = NULL, ...)
| margin | Float, margin term in the loss definition. |
|---|---|
| name | Optional name for the op. |
| ... | additional parameters to pass |
lifted_loss: tf$float32 scalar.
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