Giters

zijian-hu / SimPLE

Code for the paper: "SimPLE: Similar Pseudo Label Exploitation for Semi-Supervised Classification"

Home Page:https://arxiv.org/abs/2103.16725

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Some questions about the reported results

QiushiYang opened this issue · comments

commented

Thanks a lot for your released codes and your excellent work. I have some questions about the reported results in your paper.
(1) Does all reported results in the paper obtained by the final EMA model instead of the trainable student model?
(2) Does the reported results are obtained by calculating the mean accuracy value of the latest several (e.g., 20) iterations/epochs during the training phase or just select the best accuracy of someone iteration/epoch, or any other protocols?

commented

(1) Does all reported results in the paper obtained by the final EMA model instead of the trainable student model?

Yes, all results are obtained from the teacher (EMA) model.

(2) Does the reported results are obtained by calculating the mean accuracy value of the latest several (e.g., 20) iterations/epochs during the training phase or just select the best accuracy of someone iteration/epoch, or any other protocols?

As mentioned in our paper (page 6):

To report test accuracy, we take the checkpoint with the highest validation accuracy and report its test accuracy.

Notice that the validation set is split from the full training set, not from the test set. In other words, our labeled training set, unlabeled training set, and validation set are 3 mutually exclusive subsets of the original training dataset.

commented

Got it. Thank you so much for your kindly helps!