AI Paper Atlas

Research map / Label-efficient and robust learning / Test-time adaptation

SODA: Robust Training of Test-Time Data Adaptors

Zige Wang, Yonggang Zhang, Zhen Fang, Long Lan, Wenjing Yang, Bo Han

NeurIPS 2023 · Label-efficient and robust learning › Test-time adaptation · approach: Realistic and stable test-time adaptation

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
Use SODA instead of leaving the model unadaptedResNet-50 · CIFAR-10-Caccuracy (%): 72.39 -> 82.55Improved
Use SODA instead of leaving the model unadaptedResNet-50 · CIFAR-100-Caccuracy (%): 41.41 -> 52.41Improved
Use SODA instead of leaving the model unadaptedResNet-50 · ImageNet-Caccuracy (%): 31.36 -> 42.14Improved
Use first-order gradients rather than ZOO, with SODA-R training settingsSODA · CIFAR-10-Caccuracy (%): 82.55 -> 88.39Improved
Increase data-adaptor depth from 0 to 3 ResNet blocks (ablation)SODA · CIFAR-10-Caccuracy (%): 82.55 -> 78.65Hurt
Increase data-adaptor depth from 0 to 3 ResNet blocks (ablation)SODA-R · CIFAR-10-Caccuracy (%): 86.07 -> 88.51Improved
Add perturbation regularization (ablation)SODA · CIFAR-10-Caccuracy (%): 82.55 -> 73.40Hurt
Add perturbation regularization (ablation)SODA-R · CIFAR-100-Caccuracy (%): 58.11 -> 60.31Improved
Use Adam instead of SGD (ablation)SODA · CIFAR-100-Caccuracy (%): 52.41 -> 49.75Hurt
Use Adam instead of SGD (ablation)SODA-R · CIFAR-100-Caccuracy (%): 58.32 -> 60.31Improved
Add dropout layers (ablation)SODA · CIFAR-100-Caccuracy (%): 43.96 -> 5.47Hurt
Add dropout layers (ablation)SODA-R · CIFAR-100-Caccuracy (%): 58.27 -> 60.31Improved

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