Research map / Adversarial robustness and security / Certified and empirical robustness
Noise Tolerance of Distributionally Robust Learning
Ramzi Dakhmouche, Ivan Lunati, Hossein Gorji
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 |
|---|---|---|---|
| train with WBM (Wasserstein batch matching) loss instead of MSE under noisy data | convolutional neural operator · Navier-Stokes and wave equation operator learning, noisy train/test | test MAE | Improved |
| train with WBM instead of WDRL under 30% Cauchy-corrupted training data | Navier-Stokes operator learning · Navier-Stokes (Cauchy noise) | test MAE | Improved |
| use WBM instead of CVaR and Chi-Sq DRO | Navier-Stokes operator learning · Navier-Stokes (Gaussian noise) | test MAE | Improved |
| use WBM instead of MSE for TSMixer forecasting | TSMixer · ETDataset electric load forecasting (noisy) | test MAE | Improved |
| use WBM instead of WDRL for training time | WDRL · Navier-Stokes operator learning (training time, no tuning) | execution time (s): 3476.27 -> 2160.92 | Improved |
| WBM with multiplicative noise | WBM · Navier-Stokes (multiplicative noise) | relative MAE | No clear effect |
| WBM with strongly correlated noise | WBM · Navier-Stokes (correlated noise) | relative error | Hurt |
| WBM in low-noise regime vs MSE | WBM · low-noise regime | test error | Hurt |
| vary batch size for WBM (up to 80) (ablation) | WBM · Navier-Stokes (noise level 1) | relative test error | No clear effect |
Closest papers by meaning
- Universal generalization guarantees for Wasserstein distributionally robust models (ICLR 2025 · similarity 0.90)
- Provable Robust Overfitting Mitigation in Wasserstein Distributionally Robust Optimization (ICLR 2025 · similarity 0.90)
- A Robust Learning Approach for Regression Models Based on Distributionally Robust Optimization (JMLR 2018 · similarity 0.90)
- Outlier-Robust Wasserstein DRO (NeurIPS 2023 · similarity 0.90)
- A Unified Wasserstein Distributional Robustness Framework for Adversarial Training (ICLR 2022 · similarity 0.89)
- Outlier-Robust Distributionally Robust Optimization via Unbalanced Optimal Transport (NeurIPS 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.
