Research map / Label-efficient and robust learning / Recognition and long-tail
C-Mixup: Improving Generalization in Regression
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 |
|---|---|---|---|
| sample mixing pairs by label similarity instead of uniformly (C-Mixup vs mixup) | mixup · Exchange-Rate | RMSE: 0.0239 -> 0.0203 | Helped |
| sample mixing pairs by label similarity instead of uniformly | mixup · Airfoil | RMSE: 3.730 -> 2.717 | Helped |
| apply vanilla mixup directly to regression | ERM · Airfoil | RMSE: 2.901 -> 3.730 | Hurt |
| add C-Mixup sampling to MetaMix | MetaMix · ShapeNet1D | MSE: 4.275 -> 4.024 | Helped |
| add C-Mixup sampling to MetaMix | MetaMix · PASCAL3D | MSE: 2.135 -> 1.995 | Helped |
| use C-Mixup for out-of-distribution robustness vs ERM | ERM · PovertyMap worst-domain | R: 0.50 -> 0.53 | Helped |
| use C-Mixup vs ERM on RCF-MNIST subpopulation shift | ERM · RCF-MNIST | RMSE: 0.162 -> 0.146 | Helped |
| replace label distance with input feature distance for pair sampling (ablation) | C-Mixup · Exchange-Rate | RMSE: 0.0203 -> 0.0212 | Hurt |
| replace label distance with hidden representation distance (ablation) | C-Mixup · ShapeNet1D | MSE: 4.024 -> 4.202 | Hurt |
| replace label distance with hidden representation distance (ablation) | C-Mixup · DTI | avg R: 0.498 -> 0.483 | Hurt |
| use C-Mixup-batch (batch-level distances) instead of full-dataset distances (ablation) | C-Mixup · Exchange-Rate | RMSE: 0.0203 -> 0.0205 | No clear effect |
| add C-Mixup sampling to CutMix | CutMix · RCF-MNIST | RMSE: 0.194 -> 0.186 | Helped |
Closest papers by meaning
- mixup: Beyond Empirical Risk Minimization (ICLR 2018 · similarity 0.92)
- Global Mixup: Eliminating Ambiguity with Clustering (AAAI 2023 · similarity 0.91)
- Over-Training with Mixup May Hurt Generalization (ICLR 2023 · similarity 0.91)
- Selective Mixup Helps with Distribution Shifts, But Not (Only) because of Mixup (ICML 2024 · similarity 0.89)
- Pushing Boundaries: Mixup's Influence on Neural Collapse (ICLR 2024 · similarity 0.88)
- Catch-Up Mix: Catch-Up Class for Struggling Filters in CNN (AAAI 2024 · similarity 0.87)
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.