Research map / Label-efficient and robust learning / Domain adaptation and generalization
Feature-Critic Networks for Heterogeneous Domain Generalization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use Feature-Critic instead of the strongest listed baseline | Visual Decathlon held-out targets with SVM · Visual Decathlon held-out targets | average accuracy (%): 41.28 -> 42.29 | Helped |
| Use Feature-Critic instead of Reptile | Visual Decathlon held-out targets with SVM · Visual Decathlon held-out targets | VD-Score: 324 -> 344 | Helped |
| Use Feature-Critic instead of CrossGrad | Visual Decathlon held-out targets with KNN · Visual Decathlon held-out targets | average accuracy (%): 33.47 -> 34.96 | Helped |
| Use Feature-Critic rather than the strongest listed baseline | Visual Decathlon few-shot KNN · Visual Decathlon K-shot | average accuracy (%): 19.61 -> 20.21 | Helped |
| Replace set embedding with flattened covariance features in the critic (ablation) | Feature-Critic on Rotated MNIST · Rotated MNIST leave-one-domain-out | average accuracy (%): 96.39 -> 96.04 | Hurt |
| Use Feature-Critic instead of the strongest listed baseline | PACS trained on the train split · PACS train split | average accuracy (%): 70.0 -> 70.4 | Helped |
| Use Feature-Critic instead of MetaReg | PACS trained on train+validation · PACS train+validation split | average accuracy (%): 72.62 -> 72.10 | Hurt |
| Train a learned auxiliary feature-critic alongside supervised classification | Feature extractor for held-out Visual Decathlon domains · Visual Decathlon held-out targets | average accuracy (%): 40.27 -> 42.29 | Helped |
Closest papers by meaning
- Adaptive Methods for Real-World Domain Generalization (CVPR 2021 · similarity 0.91)
- Failure Modes of Domain Generalization Algorithms (CVPR 2022 · similarity 0.90)
- Generalizing Across Domains via Cross-Gradient Training (ICLR 2018 · similarity 0.90)
- In Search of Lost Domain Generalization (ICLR 2021 · similarity 0.89)
- A General Representation-Based Approach to Multi-Source Domain Adaptation (ICML 2025 · similarity 0.88)
- Towards Generalizing to Unseen Domains with Few Labels (CVPR 2024 · similarity 0.88)
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.