Maximal Domain Independent Representations Improve Transfer Learning

arXiv:2306.00262v2 Announce Type: replace
Abstract: State of the art domain adaptation involves the creation of (1) a domain independent representation (DIRep) trained so that from that representation it is not possible to determine whether the input is from the source domain or the target and (2) a domain dependent representation (DDRep). The original input can then be reconstructed from those two representations. The classifier is trained only on source images using the DIRep. We show that information useful only in …

abstract art arxiv cs.cv domain domain adaptation independent representation state state of the art transfer transfer learning type