Research map / Learning theory and algorithms / Prediction and decision losses
Distributionally Robust Set Representation Learning Under Inference-Time Element Corruption
Yankai Chen, Hanrong Zhang, Bowei He, Philip S. Yu, Xue Liu
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 SW-DRSO robust training instead of ERM (ablation) | SW-DRSO set encoder · LiveJ | Recall@10 (overall): 72.58 -> 74.86 | Improved |
| add SW-DRSO robust training instead of ERM (ablation) | SW-DRSO set encoder · LiveJ severe | Recall@10: 70.96 -> 73.25 | Improved |
| add sliced-Wasserstein encoder f_sw instead of mean pooling (ablation) | SW-DRSO · LiveJ | Recall@10 (overall): 69.55 -> 74.86 | Improved |
| add Wasserstein barycenter synthesis instead of random in-batch embeddings (ablation) | SW-DRSO · LiveJ | Recall@10 (overall): 72.92 -> 74.86 | Improved |
| use sliced Wasserstein instead of Sinkhorn Wasserstein (ablation) | SW-DRSO · LiveJ | training time per epoch (min): 21.5 -> 15.6 | Improved |
| use sliced Wasserstein instead of Sinkhorn Wasserstein (ablation) | SW-DRSO · LiveJ | Recall@10 (overall): 74.81 -> 74.86 | No clear effect |
| use barycentric adversary instead of discrete adversary (ablation) | SW-DRSO · LiveJ | Recall@10 (overall): 74.45 -> 74.86 | Improved |
| use barycentric adversary instead of discrete adversary (ablation) | SW-DRSO · LiveJ | training time per epoch (min): 18.4 -> 15.6 | Improved |
| increase random combinatorial search rounds m from 2 to 10 (ablation) | RCS · LiveJ | Recall@10: 69.45 -> 71.13; time 21.7 -> 27.4 | Mixed |
| use barycentric adversary instead of random combinatorial search (ablation) | SW-DRSO · LiveJ | Recall@10: 71.13 -> 74.86 | Improved |
| use SW-DRSO instead of WDRO | DRO variant · LiveJ | Recall@10 (overall): 74.72 -> 74.86 | Improved |
| use SW-DRSO instead of KL-DRO | DRO variant · LiveJ | training time per epoch (min): 14.2 -> 15.6 | Hurt |
Closest papers by meaning
- Adaptive Robust Evidential Optimization For Open Set Detection from Imbalanced Data (ICLR 2023 · similarity 0.86)
- Robust Generalization Against Photon-Limited Corruptions via Worst-Case Sharpness Minimization (CVPR 2023 · similarity 0.84)
- Modeling the Second Player in Distributionally Robust Optimization (ICLR 2021 · similarity 0.84)
- Wasserstein Distributionally Robust Optimization through the Lens of Structural Causal Models and Individual Fairness (NeurIPS 2024 · similarity 0.84)
- Distributionally Robust Optimization with Bias and Variance Reduction (ICLR 2024 · similarity 0.83)
- Distributionally Robust Token Optimization in RLHF (NeurIPS 2026 · similarity 0.83)
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.
