AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Adversarial Partial Domain Adaptation by Cycle Inconsistency

Kun-Yu Lin, Jiaming Zhou, Yukun Qiu, Wei-Shi Zheng

ECCV 2022 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Adversarial domain adaptation · 10 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add cycle-inconsistency sample weighting to DANN (ablation)DANN · Office-Home PDAavg accuracy: 55.88 -> 76.72Improved
add center losses (ablation)Ours-Cycle · Office-Home PDAavg accuracy: 76.72 -> 77.72Improved
use soft nearest neighbor instead of hard prototype search (ablation)Ours-Cycle · Office-Home PDAavg accuracy: 77.44 -> 77.72Improved
add cycle-back transformation (ablation)Ours-Src2Trg · Office-Home PDAavg accuracy: 76.33 -> 77.72Improved
add mixed strategy with source-to-target term (ablation)Ours-Cycle · Office-Home PDAavg accuracy: 77.72 -> 78.83Improved
initialize with source-only modelIWAN · Office-Home PDAavg accuracy: 69.87 -> 73.04Improved
initialize with source-only modelETN · Office-Home PDAavg accuracy: 70.45 -> 67.65Mixed
initialize with source-only modelETN · VisDA-2017 PDAavg accuracy: 66.84 -> 64.15Hurt
apply DANN adversarial alignment with no weightingsource-only model · Office-Home PDAavg accuracy: 61.77 -> 55.88Hurt
use cycle weighting vs prior weighted adversarial methodsOurs · VisDA-2017 PDARe.->Sy. accuracy: 78.18 -> 86.50Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy