AI Paper Atlas

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

S3VAADA: Submodular Subset Selection for Virtual Adversarial Active Domain Adaptation

Harsh Rangwani, Arihant Jain, Sumukh K Aithal, R. Venkatesh Babu

ICCV 2021 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Adversarial domain adaptation · 29 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 gradient clipping (norm 1) to VAADA training (ablation)VAADA · Office-31 Webcam->Amazonaccuracy: above 10% per cycleImproved
do not lower backbone learning rate (same rate as head) (ablation)VAADA · Office-31 Webcam->AmazonaccuracyImproved
use VAADA instead of VADA/DANN as adaptation procedure (ablation)S3VAADA sampling · Office-31 Webcam->Amazonaccuracy: >= 5% per cycleImproved
combine uncertainty, diversity and representativeness (alpha=0.5, beta=0.3) instead of uncertainty alone or diversity alone (ablation)S3VAADA sampling · Office-31 Webcam->AmazonaccuracyImproved
use S3VAADA sampling instead of other sampling methods under VAADA trainingVAADA · Office-31 Webcam->Amazonaccuracy: 9% with 45 labeled samplesImproved
use S3VAADA instead of other methods on VisDA-18S3VAADA · VisDA-18accuracy: 7% averaged across cyclesImproved
use S3VAADA instead of other methods on easier Art->Product shiftS3VAADA · Office-Home Art->ProductaccuracyNo clear effect
use S3VAADA instead of other methods on harder Office-Home shiftsS3VAADA · Office-Home Art->Clipartaccuracy: 3%-5%Improved
increase number of active rounds beyond 15S3VAADA · Office-31 Webcam->Amazonaccuracy: 89% at 20 roundsNo clear effect
use very small labeling budget of 25 instead of >45 (ablation)S3VAADA · Office-31 Webcam->AmazonaccuracyHurt
use S3VAADA instead of MME* semi-supervised DA with random labelingS3VAADA · Office-31accuracyImproved

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