Research map / Learning theory and algorithms / Optimal transport
Tikhonov Regularization is Optimal Transport Robust under Martingale Constraints
Jiajin Li, Sirui Lin, Jose Blanchet, Viet Anh Nguyen
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 |
|---|---|---|---|
| Martingale DRO instead of ERM | ResNet · CIFAR-10 PGD epsilon = 0.08 | Top-1 accuracy (%): 70.20 -> 80.86 | Improved |
| Martingale DRO instead of ERM | ResNet · CIFAR-10 PGD epsilon = 0.04 | Top-1 accuracy (%): 77.50 -> 83.25 | Improved |
| Martingale DRO instead of ERM | ResNet · CIFAR-10 PGD epsilon = 0 | Top-1 accuracy (%): 84.16 -> 85.48 | Improved |
| Martingale DRO instead of DRO | ResNet · CIFAR-10 PGD epsilon = 0.08 | Top-1 accuracy (%): 80.68 -> 80.86 | Improved |
| Martingale DRO instead of DRO | ResNet · CIFAR-10 PGD epsilon = 0.04 | Top-1 accuracy (%): 82.87 -> 83.25 | Improved |
| Martingale DRO instead of DRO | ResNet · CIFAR-10 PGD epsilon = 0 | Top-1 accuracy (%): 84.02 -> 85.48 | Improved |
| Martingale DRO instead of Jacobian regularization | ResNet · CIFAR-10 PGD epsilon = 0.08 | Top-1 accuracy (%): 73.85 -> 80.86 | Improved |
| Increasing PGD perturbation from 0 to 0.08 for ERM | ERM · CIFAR-10 | Top-1 accuracy (%): 84.16 -> 70.20 | Hurt |
| Increasing PGD perturbation from 0 to 0.08 for Martingale DRO | Martingale DRO · CIFAR-10 | Top-1 accuracy (%): 85.48 -> 80.86 | Hurt |
| Small martingale violation size epsilon (ablation) | Martingale DRO · MNIST | test error | Hurt |
| Large martingale violation size epsilon (ablation) | Martingale DRO · MNIST | test error | No clear effect |
Closest papers by meaning
- Generalization Bounds for (Wasserstein) Robust Optimization (NeurIPS 2021 · similarity 0.89)
- Outlier-Robust Distributionally Robust Optimization via Unbalanced Optimal Transport (NeurIPS 2024 · similarity 0.89)
- Regularized Optimal Transport is Ground Cost Adversarial (ICML 2020 · similarity 0.89)
- Universal generalization guarantees for Wasserstein distributionally robust models (ICLR 2025 · similarity 0.88)
- Distributionally Robust Optimization with Bias and Variance Reduction (ICLR 2024 · similarity 0.87)
- Wasserstein Distributionally Robust Optimization through the Lens of Structural Causal Models and Individual Fairness (NeurIPS 2024 · similarity 0.86)
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.
