Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain Generalization Guided by Gradient Signal to Noise Ratio of Parameters
Mateusz Michalkiewicz, Masoud Faraki, Xiang Yu, Manmohan Chandraker, Mahsa Baktashmotlagh
This paper tried a common change: Add dropout
add GSNR-guided dropout to SSAN · OCIM average HTER · HTER: 13.21 -> 11.02 Improved
What 795 accepted papers report for this change
- 549 improved
- 87 no clear effect
- 12 mixed
- 147 hurt
- Don't Drop Dropout: Optimizing Layer Sparsity for Efficient LLM Training and Inference ICML 2026 · apply layer dropout to whole transformer blocks instead of attention and FFN sub-blocks separately · 20 TPP val loss, 271M · val loss: 2.377 -> 2.367Improved
- Quantifying the Uncertainty of Foundation Models with Singular Value Ensembles ICML 2026 · use MC Dropout · vision and SST-2 · accuracy: 86.3 -> 76.6 on Flowers102Hurt
- Motion Planning in Compressed Representation Spaces ICML 2026 · add nested dropout during training · WOMD guided generation · goal FDE: 4.145 -> 1.104Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
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 changed | Where | Measured | Effect |
|---|---|---|---|
| mask by GSNR instead of random Bernoulli in DropBlock (ablation) | DropBlock · PACS | average accuracy: 79.2 -> 84.43 | Improved |
| mask by GSNR instead of gradient magnitude in DropBlock (ablation) | DropBlock · PACS | average accuracy: 83.15 -> 84.43 | Improved |
| mask by GSNR instead of gradient magnitude in SpatialDropout (ablation) | SpatialDropout · PACS | average accuracy: 83.22 -> 83.31 | No clear effect |
| mask by GSNR instead of gradient magnitude in RSC (ablation) | RSC · PACS | average accuracy: 83.90 -> 84.38 | Improved |
| meta-learn dropout ratios instead of grid search (ablation) | DB + GSNR · PACS | average accuracy: 85.37 -> 85.84 | Improved |
| learn dropout ratios by meta-learning instead of fixed ratio (ablation) | DB + GSNR · PACS | average accuracy: 84.43 -> 85.84 | Improved |
| add CutOut, MixUp, CutMix or Manifold Mixup to baseline (ablation) | ResNet18 baseline · PACS | average accuracy: 79.5 -> 76.8 (CutMix) | Hurt |
| add GSNR-guided dropout to SSAN | SSAN · OCIM average HTER | HTER: 13.21 -> 11.02 | Improved |
| add GSNR-guided dropout to EPCR | EPCR · OCIM average HTER | HTER: 11.36 -> 10.47 | Improved |
| GSNR dropout vs RSC on Office-Home | RSC · Office-Home | average accuracy: 63.12 -> 65.28 | Improved |
| add GSNR-guided dropout to SSAN on the O&C&I to M protocol | SSAN · OCIM O&C&I to M AUC | AUC: 97.50 -> 95.50 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Style Neophile — Style Neophile: Constantly Seeking Novel Styles for Domain Generalization (CVPR 2022) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Test-Time Domain Generalization for Face Anti-Spoofing (CVPR 2024 · similarity 0.86)
- Gradient Alignment for Cross-Domain Face Anti-Spoofing (CVPR 2024 · similarity 0.86)
- Single-Side Domain Generalization for Face Anti-Spoofing (CVPR 2020 · similarity 0.86)
- Meta Dropout: Learning to Perturb Latent Features for Generalization (ICLR 2020 · similarity 0.86)
- Regularized Fine-Grained Meta Face Anti-Spoofing (AAAI 2020 · similarity 0.86)
- Sharpness-Aware Model-Agnostic Long-Tailed Domain Generalization (AAAI 2024 · similarity 0.84)
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.
