DARN implementation and training strategy

[pbamotra]

Hi authors,

Can you please explain what was your methodology of training DARN network? DARN has n-branches for each attribute group and expects to have one label per class but in deep fashion data, many a times, multiple attributes belonging to same group get activated (multi-label per class). How are training DARN in that case?