szha commented on a change in pull request #12750: [MXNET 1030] Cosine Embedding Loss
URL: https://github.com/apache/incubatormxnet/pull/12750#discussion_r228362925
##########
File path: python/mxnet/gluon/loss.py
##########
@@ 767,3 +767,69 @@ def hybrid_forward(self, F, pred, target, sample_weight=None, epsilon=1e08):
loss += stirling_factor
loss = _apply_weighting(F, loss, self._weight, sample_weight)
return F.mean(loss)
+
+
+class CosineEmbeddingLoss(Loss):
+ r"""For a target label 1 or 1, vectors target and pred, the function computes the cosine
distance
+ between the vectors. This can be interpretted as how similar/dissimilar two input vectors
are.
+
+ .. math::
+
+ L = \sum_i \begin{cases} 1  {cos\_sim({input1}_i, {input2}_i)} & \text{ if }
{label}_i = 1\\
+ {cos\_sim({input1}_i, {input2}_i)} & \text{ if } {label}_i =
1 \end{cases}\\
+ cos\_sim(input1, input2) = \frac{{input1}_i.{input2}_i}{{input1}_i.{input2}_i}
+
+ `input1`, `input2` can have arbitrary shape as long as they have the same number of elements.
+
+ Parameters
+ 
+ weight : float or None
+ Global scalar weight for loss.
+ batch_axis : int, default 0
+ The axis that represents minibatch.
+ margin : float
+ Margin of separation between correct and incorrect pair.
+
+
+ Inputs:
+  **input1**: a tensor with arbitrary shape
+  **input2**: another tensor with same shape as pred to which input1 is
+ compared for similarity and loss calculation
+  **sample_weight**: elementwise weighting tensor. Must be broadcastable
Review comment:
This needs to be added to hybrid_forward. Could you also put this after `label`?

This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
