Closed sehyungp92 closed 3 years ago
@sehyungp92 did you reproduce the results of the multiclass_ood detector now? When I run the code the first time, the AUROC is only 0.54718556.
Did the network train properly? If it has reasonable classification accuracy its AUROC shouldn't be nearly that low.
On Mon, Jul 13, 2020 at 7:03 PM Fourk_tx notifications@github.com wrote:
@sehyungp92 https://github.com/sehyungp92 did you reproduce the results of the multiclass_ood detector now? When I run the code the first time, the AUROC is only 0.54718556.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/hendrycks/ss-ood/issues/14#issuecomment-657922144, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACZBITUGO6S4WY2PTREOTGTR3O4GZANCNFSM4OD4MJHA .
Hi, @hendrycks , I thought it was not trained properly, but I did not change the hyper-parameters. Can you share your hyper-parameters to us?
Thanks for your reply.
It looks like the default hyperparamter is to train for 5 epochs. That should probably be 100 (sorry this code was added long after the experiments were run).
On Mon, Jul 13, 2020 at 7:21 PM Fourk_tx notifications@github.com wrote:
Hi, @hendrycks https://github.com/hendrycks , I thought it was not trained properly, but I did not change the hyper-parameters. Can you share your hyper-parameters to us?
Thanks for your reply.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/hendrycks/ss-ood/issues/14#issuecomment-657926785, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACZBITV46GNTMHMGMN4YG3DR3O6I7ANCNFSM4OD4MJHA .
Firstly many thanks for sharing your work.
I found the paper very interesting and wanted to see if I could reproduce some of the results, especially regarding the one-class OOD detector on CIFAR-10 using OE. I was just wondering if this would be made available at any point.
Kind regards, Se