AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

Modeling the Second Player in Distributionally Robust Optimization

Paul Michel, Tatsunori Hashimoto, Graham Neubig

ICLR 2021 · Learning theory and algorithms › Prediction and decision losses · approach: Distributionally robust optimization · 3 citations (OpenAlex)

Read the paper

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 changedWhereMeasuredEffect
Use P-DRO instead of ERM without group informationBiasedSST · BiasedSST testrobust accuracy: 2.15 ± 0.97 -> 34.98 ± 9.39Improved
Use P-DRO instead of NonParamBiasedSST · BiasedSST testrobust accuracy: 28.11 ± 2.16 -> 34.98 ± 9.39Improved
Use P-DRO instead of ERM without group informationDWMW17 · DWMW17 test, no group informationrobust accuracy: 53.19 ± 1.70 -> 69.06 ± 1.70Improved
Use P-DRO instead of NonParam without group informationDWMW17 · DWMW17 test, no group informationrobust accuracy: 54.13 ± 1.14 -> 69.06 ± 1.70Improved
Use P-DRO instead of ERM without group informationFDCL18 · FDCL18 test, no group informationrobust accuracy: 19.57 ± 7.00 -> 30.25 ± 10.13Improved
Use P-DRO instead of ERM when validation group annotations are availableDWMW17 · DWMW17 test, group information on validation setrobust accuracy: 53.15 ± 0.87 -> 63.05 ± 4.25Improved
Use P-DRO instead of ERM when validation group annotations are availableFDCL18 · FDCL18 test, group information on validation setrobust accuracy: 34.07 ± 3.20 -> 47.61 ± 4.53Improved
Add the KL constraint to Minmax stopping (ablation)P-DRO · BiasedSST validation, stopping criterionrobust accuracy: 25.22 ± 13.01 -> 31.30 ± 10.07Improved
Use Greedy-Minmax stopping with the KL constraint (ablation)P-DRO · BiasedSST validation, stopping criterionrobust accuracy: 31.30 ± 10.07 -> 32.17 ± ± 11.20Improved
Replace Average stopping with Minmax stopping (ablation)P-DRO · BiasedSST validation, stopping criterionrobust accuracy: 0.00 ± 0.00 -> 25.22 ± 13.01Improved
Use KL-constrained Minmax rather than Average for hyperparameter selection (ablation)P-DRO · BiasedSST validation, hyperparameter selectionrobust accuracy: 31.03 ± 12.16 -> 35.65 ± 11.47Improved
Add the approximate KL constraint to bare P-DRO (ablation)P-DRO toy task · BiasedSST toy taskrobust accuracy: 49.85 ± 0.10 -> 58.41 ± 9.25Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy