Reproducing the Results

[mgholamikn]

Hello,

Thanks a lot for sharing your code. I was trying to reproduce your results on CIFAR10 and got the following results:

The results of LogME and Leep are exactly similar to numbers in SFDA paper. However, the results of SFDA is not similar. I also got Kendall tau of 0.518 by SFDA on the VOC2007 that is way lower than the numbers in SFDA paper. Could you please let me know if there is any point that I need to consider to reproduce the results?

[wqshao126]

Thanks for your interest. We have reimplemented the case following the readme file. It reports the same number with the paper. Could you provide more details about your implementation? such as a in eq.3, data loader (we use all_loader in all cases).

[mgholamikn]

Thanks for the reply. I didn't change anything in the code. I just downloaded the code and data following the instructions