Research map / Learning theory and algorithms / Prediction and decision losses
Gen-DFL: Decision-Focused Generative Learning for Robust Decision Making
Prince Zizhuang Wang, Shuyi Chen, Jinhao Liang, Ferdinando Fioretto, Shixiang 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 |
|---|---|---|---|
| use Gen-DFL instead of SPO+ | Gen-DFL · Shortest Path Deg-8 (sigma=20) | average relative regret: 81.78 -> 13.36 | Improved |
| use Gen-DFL instead of Diff-DRO on Portfolio | Gen-DFL · Portfolio (sigma=20) | average relative regret: up to 58.5% | Improved |
| use Gen-DFL instead of SPO+ on Portfolio | Gen-DFL · Portfolio (sigma=20) | average relative regret: up to 48.5% | Improved |
| use Gen-DFL instead of SPO+ and Diff-DRO on Knapsack Deg-2 | Gen-DFL · Knapsack Deg-2 (sigma=20) | average relative regret: 19.6% and 10.3% | Improved |
| use Gen-DFL instead of SPO+ on Knapsack Deg-6 | Gen-DFL · Knapsack Deg-6 (sigma=20) | average relative regret: 17.45 -> 17.91 | Hurt |
| use Gen-DFL instead of Diff-DRO on COVID Resource | Gen-DFL · COVID Resource (sigma=20) | average relative regret: 16.41 -> 16.86 | Hurt |
| add decision-cost weight beta > 0 to the generative loss (ablation) | Gen-DFL · Portfolio, risk-sensitive evaluation | regret | Improved |
| train with smaller alpha (0.5 instead of 1.0) (ablation) | Gen-DFL · Portfolio, risk-sensitive evaluation | CVaR regret | Improved |
| increase number of generated samples in SAA (ablation) | Gen-DFL · Portfolio, risk-sensitive evaluation | regret | Improved |
| use conditional normalizing flow instead of VAE as generative model (ablation) | Gen-DFL · generative model choice | regret | Improved |
| increase noise variance sigma (40 to 100) (ablation) | Gen-DFL · Portfolio | regret | No clear effect |
Closest papers by meaning
- Online Decision-Focused Learning (ICLR 2026 · similarity 0.86)
- Decision-Focused Learning without Decision-Making: Learning Locally Optimized Decision Losses (NeurIPS 2022 · similarity 0.86)
- DFF: Decision-Focused Fine-Tuning for Smarter Predict-Then-Optimize with Limited Data (AAAI 2025 · similarity 0.85)
- Feasibility-Aware Decision-Focused Learning for Predicting Parameters in the Constraints (NeurIPS 2025 · similarity 0.85)
- Solver-Free Decision-Focused Learning for Linear Optimization Problems (NeurIPS 2025 · similarity 0.84)
- Locally Convex Global Loss Network for Decision-Focused Learning (AAAI 2025 · similarity 0.82)
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.
