Thanks very much for your reply.
Seen from Figure for the training and testing loss in your original paper, you dumped the testing data during the training process. So the testing data is no longer unseen for the model, if I understand correctly.
Best
Training and testing datasets have no samples in common. Testing dataset only includes unseen samples in the training phase. I'm not so sure what your problem is .
https://uploads.disquscdn.c...
Thanks for your patience.
I was just wondering, based on the figure in your paper, the testing data was used at the beginning of training?
Thanks in advance
