Research map / Optimization / Optimizers
Gradient Flow Sampler-based Distributionally Robust Optimization
Zusen Xu, Jia-Jie Zhu
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 |
|---|---|---|---|
| add birth-death (Fisher-Rao) reaction to Wasserstein gradient flow sampler (ablation) | WGF-DRO · MNIST LeNet-5 PGD-L2 (delta=0.075) | test error (%): 21.57 -> 20.81 | Improved |
| add birth-death reaction to WGF-DRO (ablation) | WGF-DRO · CIFAR-10 ResNet-18 PGD-L2 (delta=0.015) | test error (%): 27.70 -> 26.73 | Improved |
| add birth-death reaction to WGF-DRO (ablation) | WGF-DRO · CIFAR-10 ResNet-18 PGD-Linf (delta=0.015) | test error (%): 35.02 -> 34.06 | Improved |
| add Langevin noise (SDE) to WRM's ODE inner step, i.e. entropy regularization (ablation) | WRM · MNIST LeNet-5 PGD-L2 (delta=0.075) | test error (%): 29.22 -> 21.57 | Improved |
| replace deterministic ODE with SDE in the inner sampler (ablation) | WRM · CIFAR-10 ResNet-18 PGD-L2 (delta=0.015) | test error (%): 71.24 -> 27.70 | Improved |
| use WRM instead of plain SAA training | SAA · MNIST LeNet-5 | test error (%): 10.80 -> 11.66 (delta=0.05, L2) | No clear effect |
| use WRM instead of plain SAA training | SAA · CIFAR-10 ResNet-18 PGD-L2 (delta=0.015) | test error (%): 78.32 -> 71.24 | Improved |
| use WFR-DRO instead of Dual Sinkhorn DRO | Dual · MNIST LeNet-5 PGD-L2 (delta=0.075) | test error (%): 24.20 -> 20.81 | Improved |
| use WFR-DRO instead of Dual Sinkhorn DRO | Dual · CIFAR-10 ResNet-18 PGD-Linf (delta=0.015) | test error (%): 39.74 -> 34.06 | Improved |
| use WFR-DRO instead of SAA on clean data | SAA · CIFAR-10 ResNet-18 (delta=0.00) | test error (%): 6.35 -> 11.10 | Hurt |
| use WFR-DRO instead of SAA on clean data | SAA · MNIST LeNet-5 (delta=0.00) | test error (%): 0.92 -> 0.92 | No clear effect |
| use Stein variational gradient flow sampler (SVG-DRO) | GF-DRO · convex feature-extracted CIFAR-10 | robustness | Hurt |
Closest papers by meaning
- Distributionally Robust Optimization with Bias and Variance Reduction (ICLR 2024 · similarity 0.90)
- Large-Scale Non-convex Stochastic Constrained Distributionally Robust Optimization (AAAI 2024 · similarity 0.89)
- Outlier-Robust Distributionally Robust Optimization via Unbalanced Optimal Transport (NeurIPS 2024 · similarity 0.88)
- Drago: Primal-Dual Coupled Variance Reduction for Faster Distributionally Robust Optimization (NeurIPS 2024 · similarity 0.88)
- Wasserstein Distributionally Robust Optimization through the Lens of Structural Causal Models and Individual Fairness (NeurIPS 2024 · similarity 0.88)
- Distributionally Robust Performative Optimization (NeurIPS 2025 · 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.
