search

WuChenshen / MeRGAN

MIT License
60 stars 17 forks source link

unreproducible classification accuracy for 10 tasks #4

Closed qichenglao closed 5 years ago

qichenglao commented 5 years ago

Hi, thanks for sharing the code. I am trying to reproduce your work. For both datasets (mnist and svhn), the classification accuracy for the 5 tasks are the same as in the paper. However, for the 10 tasks, I only got ~0.94 for the mnist and ~0.63. for svhn. Is there anything missing in the code in order to reproduce your results in the paper?

qichenglao commented 5 years ago

Sorry, it is reproduced by increasing iteration to 8001 for mnist.

liuziye1997 commented 4 years ago

Why the accuracy of my running results is close to zero for the previous categories？ That is, each prediction is predicted as the last category.