Research map / Label-efficient and robust learning / Domain adaptation and generalization
Efficient Domain Generalization via Common-Specific Low-Rank Decomposition
Vihari Piratla, Praneeth Netrapalli, Sunita Sarawagi
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use CSD instead of MASF | LeNet · LipitK, 25 domains | accuracy: 78.5 (0.5) -> 77.6 (0.4) | Hurt |
| Use CSD instead of MASF | LeNet · LipitK, 76 domains | accuracy: 85.9 (0.3) -> 87.3 (0.4) | Improved |
| Use CSD instead of ERM | Speech classifier · Speech utterances, 100 domains | accuracy: 80.0 (.1) -> 81.4 (.4) | Improved |
| Use CSD instead of ERM | ResNet-18 · Rotated MNIST, in-domain | accuracy: 98.3 (0.0) -> 98.4 (0.0) | No clear effect |
| Use CSD instead of ERM | ResNet-18 · Rotated MNIST, out-domain | accuracy: 93.6 (0.7) -> 94.7 (0.2) | Improved |
| Add specialized loss to common loss with orthonormality regularizer (ablation) | CSD on LipitK · LipitK, 76 domains | accuracy: 85.7 (.4) -> 87.3 (.3) | Improved |
| Add orthonormality regularizer to common and specialized losses (ablation) | CSD on LipitK · LipitK, 76 domains | accuracy: 85.8 (.6) -> 87.3 (.3) | Improved |
| Use specialized loss only instead of common loss only (ablation) | CSD on LipitK · LipitK, 76 domains | accuracy: 85.5 (.7) -> 84.4 (.2) | Hurt |
| Add orthonormality regularizer to specialized loss (ablation) | CSD on LipitK · LipitK, 76 domains | accuracy: 84.4 (.2) -> 85.3 (.1) | Improved |
| Increase domain-specific rank from 0 to 1 (ablation) | CSD · Speech utterances, 50 domains | accuracy: 72.6 (.1) -> 74.1 (.3) | Improved |
| Increase domain-specific rank from 1 to 24 (ablation) | CSD · Speech utterances, 50 domains | accuracy: 74.1 (.3) -> 72.3 (.2) | Hurt |
| Increase domain-specific rank from 1 to 4 (ablation) | CSD · Speech utterances, 200 domains | accuracy: 87.3 (.5) -> 87.5 (.1) | Improved |
Closest papers by meaning
- Learning to Balance Specificity and Invariance for In and Out of Domain Generalization (ECCV 2020 · similarity 0.90)
- Failure Modes of Domain Generalization Algorithms (CVPR 2022 · similarity 0.90)
- Exploiting Domain-Specific Features to Enhance Domain Generalization (NeurIPS 2021 · similarity 0.89)
- Learning to Optimize Domain Specific Normalization for Domain Generalization (ECCV 2020 · similarity 0.89)
- Lost Domain Generalization Is a Natural Consequence of Lack of Training Domains (AAAI 2024 · 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.
