Can you share log file of training cifar10 of asym noise?

Question

Can you share log file of training cifar10 of asym noise?

wjun0830 opened this issue 4 years ago · comments

Hi Junnan. I really appreciate your work and it helps me study a lot actually.

However, somehow, I cannot get the result as same as you mentioned in your paper.
It shows that asym 40% noise on cifar-10 should give 93% accuracy but what i see is middle of 80s.

Can you share your log file of asym0.4-cifar10?

This is image of middle of the process.

Junnan Li · Answer 1 · Thu Aug 06 2020 08:19:04 GMT+0800 (China Standard Time)

Hi, thanks for trying our code!
For asym noise, you need to set lambda_u = 0, as shown in our paper's appendix. Also, you need to finish 300 epochs to get the reported performance.

WonJun Moon · Answer 2 · Fri Aug 07 2020 08:08:28 GMT+0800 (China Standard Time)

Oh i missed that. Thanks

Chen Feng · Answer 3 · Mon Oct 26 2020 12:40:26 GMT+0800 (China Standard Time)

Hi Junnan, could you explain a bit why unlabeled part loss should be avoid in asym noise mode? Thanks so much!

Junnan Li · Answer 4 · Mon Oct 26 2020 16:33:10 GMT+0800 (China Standard Time)

Hi,
the labeled loss already contain information for the unlabeled samples because of mixup, and we found the extra regularization of the unlabeled loss to be not necessary.