Hi,
I am struggling a bit whether I fully understood the MDL training and splitting. Is the following diagram correct or am I missing something? Is Euclidean Alignment and Mixup still used during Phase 2 and 3? Does Phase 2 train a random classifier or does it finetune the classifier from phase1?
Thanks in advance!
Hi,
I am struggling a bit whether I fully understood the MDL training and splitting. Is the following diagram correct or am I missing something? Is Euclidean Alignment and Mixup still used during Phase 2 and 3? Does Phase 2 train a random classifier or does it finetune the classifier from phase1?
Thanks in advance!