You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We reproduced the experiment and found:
(1) For CUB200 data set, the acc of session 0 is 80.17%, when using the settings in the paper, i.e. the classification module is trained for 1000 epochs at a learning rate of 0.1 for the base session. It is better than the paper reported 79.60%.
So we guess that, the reported accuracy 79.60% of session 0 is obtained from classification module trained for 500 epochs.
(2) We check the acc at 500 epochs, as expected, it is 79.61%, very close to the reported.
Is the acc of session 0 for CUB200 obtained at 500 epochs?
The text was updated successfully, but these errors were encountered:
We reproduced the experiment and found:
(1) For CUB200 data set, the acc of session 0 is 80.17%, when using the settings in the paper, i.e. the classification module is trained for 1000 epochs at a learning rate of 0.1 for the base session. It is better than the paper reported 79.60%.
So we guess that, the reported accuracy 79.60% of session 0 is obtained from classification module trained for 500 epochs.
(2) We check the acc at 500 epochs, as expected, it is 79.61%, very close to the reported.
Is the acc of session 0 for CUB200 obtained at 500 epochs?
The text was updated successfully, but these errors were encountered: