Jing
Results
1
comments of
Jing
> if you read section J.3 in the paper, the code is identical Thanks for your reply! I see, after normalization the L2 loss is the same as cosine similarity