Thanks for ur excellent and laborious work!
I was looking for the code of Mixup, and got to this respository
I have a question for the data flow: In the setting of domain adaptation, the 'uda_device' means the data without labels in target domain in ur file 'train.py'L211-L212. While if using the algorightm Mixup which inherits from ERM, the method 'update' of both class donot use the unlabeled data which is exactly the 'uda_device', the original paper indicated both domains(source and target) are used to produce new data.
Is there something wrong with my understanding?
Thanks for ur work again and Looking for ur reply
Thanks for ur excellent and laborious work! I was looking for the code of Mixup, and got to this respository I have a question for the data flow: In the setting of domain adaptation, the 'uda_device' means the data without labels in target domain in ur file 'train.py'L211-L212. While if using the algorightm Mixup which inherits from ERM, the method 'update' of both class donot use the unlabeled data which is exactly the 'uda_device', the original paper indicated both domains(source and target) are used to produce new data. Is there something wrong with my understanding? Thanks for ur work again and Looking for ur reply