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
In your paper, you write: "we conduct the leave-one-domain-out evaluation. We train our model on the training splits and select the best model on the validation splits of all source domains"
But, in my opinion and according to the description in "https://github.com/facebookresearch/DomainBed", "select the best model on the validation splits of all source domains" is the "Training-domain validation" or "IIDAccuracySelectionMethod", rather than the " leave-one-domain-out evaluation".
What's your opiniton? Thanks for your reply.
The text was updated successfully, but these errors were encountered:
In your paper, you write: "we conduct the leave-one-domain-out evaluation. We train our model on the training splits and select the best model on the validation splits of all source domains"
But, in my opinion and according to the description in "https://github.com/facebookresearch/DomainBed", "select the best model on the validation splits of all source domains" is the "Training-domain validation" or "IIDAccuracySelectionMethod", rather than the " leave-one-domain-out evaluation".
What's your opiniton? Thanks for your reply.
The text was updated successfully, but these errors were encountered: