Hi, thanks for your work,
but I'm concerned about why there is no dropout in your model? Dropout improves the performance of consistent loss largely as a data augmentation policy. Is this why the performance of the Pi model is so different from that of the VAT?
Thank you
Hi, thanks for your work, but I'm concerned about why there is no dropout in your model? Dropout improves the performance of consistent loss largely as a data augmentation policy. Is this why the performance of the Pi model is so different from that of the VAT? Thank you