Re-implementation of your result about P-correct

LiJunnan1992 / DivideMix

Code for paper: DivideMix: Learning with Noisy Labels as Semi-supervised Learning

MIT License

543 stars 84 forks source link

Re-implementation of your result about P-correct #16

Closed congyang1996 closed 4 years ago

congyang1996 commented 4 years ago

Hi Junnan, so nice of your work! In your paper, you cited this work, P-correction (Yi & Wu, 2019) , and re-implemented this work, get worse result than his paper. I have done the same thing, and get even worse result than your re-implementation. Could you please share some experience in re-implementation? Thank you a lot! ^_^

LiJunnan1992 commented 4 years ago

Hi thanks for asking! I mostly follow their paper for the re-implementation, but I do observe that training could sometimes be a bit unstable. It is probably better that you check with the authors for their paper :)