DiDA: Iterative Boosting of Disentangled Synthesis and Domain Adaptation

Jinming Cao, Oren Katzir, Peng Jiang, Dani Lischinski, Daniel Cohen-Or, Changhe Tu, Yangyan Li

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Unsupervised domain adaptation aims at learning a shared model for two related domains by leveraging supervision from a source domain to an unsupervised target domain. A number of effective domain adaptation approaches rely on the ability to extract domain-invariant latent factors which are common to both domains. Extracting latent commonality is also useful for disentanglement analysis. It enables separation between the common and the domain-specific features of both domains, which can be recombined for synthesis. In this paper, we propose a strategy to boost the performance of domain adaptation and disentangled synthesis iteratively. The key idea is that by learning to separately extract both the common and the domain-specific features, one can synthesize more target domain data with supervision, thereby boosting the domain adaptation performance. Better common feature extraction, in turn, helps further improve the feature disentanglement and the following disentangled synthesis. We show that iterating between domain adaptation and disentangled synthesis can consistently improve each other on several unsupervised domain adaptation benchmark datasets and tasks, under various domain adaptation backbone models.

Original languageEnglish
Title of host publicationProceedings - 11th International Conference on Information Technology in Medicine and Education, ITME 2021
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages201-208
Number of pages8
ISBN (Electronic)9781665406796
DOIs
StatePublished - 2021
Event11th International Conference on Information Technology in Medicine and Education, ITME 2021 - Wuyishan, China
Duration: 19 Nov 202121 Nov 2021

Publication series

NameProceedings - 11th International Conference on Information Technology in Medicine and Education, ITME 2021

Conference

Conference11th International Conference on Information Technology in Medicine and Education, ITME 2021
Country/TerritoryChina
CityWuyishan
Period19/11/2121/11/21

Keywords

  • Disentanglement
  • Domain Adaptation
  • Unsupervised Learning

Fingerprint

Dive into the research topics of 'DiDA: Iterative Boosting of Disentangled Synthesis and Domain Adaptation'. Together they form a unique fingerprint.

Cite this