Research map / Learning theory and algorithms / Combinatorial and search optimization
Interior Point Solving for LP-based prediction+optimisation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 IntOpt rather than SPO | Real estate knapsack · Real estate knapsack | Regret (× 10^4): 295 -> 457 | Hurt |
| Train with IntOpt rather than SPO | Energy scheduling (0-layer) · Energy scheduling (0-layer) | Regret: 11073 -> 10774 | Helped |
| Train with IntOpt rather than SPO | Shortest path · Shortest path | Regret: 152 -> 119; 138 -> 92 | Helped |
| Add log-barrier regularization instead of squared-norm regularization (ablation) | KKT-based IntOpt on energy scheduling · Energy scheduling | Regret: 15744 -> 14365 (λ=10^-1); 174717 -> 14958 (λ=10^-3); 209595 -> 21258 (λ=10^-10) | Helped |
| Differentiate the HSD formulation rather than KKT conditions (ablation) | Log-barrier IntOpt on energy scheduling · Energy scheduling | Regret: 14365 -> 10774 (λ=10^-1); 14958 -> 14620 (λ=10^-3); 21258 -> 21594 (λ=10^-10) | Helped |
| Stop at a lower λ cut-off of 10^-3 rather than 10^-1 (ablation) | HSD log-barrier IntOpt · Energy scheduling | Regret: 10774 -> 14620 | Hurt |
| Stop at a lower λ cut-off of 10^-10 rather than 10^-1 (ablation) | HSD log-barrier IntOpt · Energy scheduling | Regret: 10774 -> 21594 | Hurt |
| Use a 1-layer rather than 0-layer network (ablation) | IntOpt · Energy scheduling | Regret: 10774 -> 11406 | Hurt |
| Use a 2-layer rather than 1-layer network (ablation) | IntOpt · Shortest path | Regret: 119 -> 92 | Helped |
Closest papers by meaning
- Solver-Free Decision-Focused Learning for Linear Optimization Problems (NeurIPS 2025 · similarity 0.85)
- Multi-Stage Predict+Optimize for (Mixed Integer) Linear Programs (NeurIPS 2024 · similarity 0.85)
- A Differentiable Interior-Point Method in Single Precision (NeurIPS 2026 · similarity 0.84)
- Principled Data Augmentation for Learning to Solve Quadratic Programming Problems (NeurIPS 2025 · similarity 0.84)
- Learning Initial Basis Selection for Linear Programming via Duality-Inspired Tripartite Graph Representation and Comprehensive Supervision (ICML 2025 · similarity 0.84)
- Learning Generalized Linear Programming Value Functions (NeurIPS 2024 · similarity 0.84)
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.