Research map / Learning theory and algorithms / Combinatorial and search optimization
Learning to Configure Separators in Branch-and-Cut
Sirui Li, Wenbin Ouyang, Max B. Paulus, Cathy Wu
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 |
|---|---|---|---|
| learn instance-aware separator configurations with L2Sep | SCIP · Standard MILP benchmarks (six classes) | relative time improvement: 42.3%; 71.9%; 28.5%; 66.2%; 72.4%; 29.4% | Improved |
| learn instance-aware separator configurations with L2Sep | SCIP · MIPLIB, NN Verification, Load Balancing | relative time improvement: 12.9%; 37.5%; 21.2% | Improved |
| learn separator configurations with one update | Gurobi · Max. Cut, Pack., Comb. Auc., Fac. Loc. | relative time improvement: 45.4%; 30.6%; 12.6%; 56.7% | Improved |
| remove configuration-space restriction (ablation) | L2Sep · Bin. Pack. and Indep. Set | relative time improvement: 42.3% -> 18.6%; 72.4% -> 38.6% | Hurt |
| use greedy restriction without the filtering criterion (ablation) | L2Sep · Bin. Pack. and Indep. Set | relative time improvement: 42.3% -> 35.8%; 72.4% -> 68.5% | Hurt |
| add a second configuration update rather than use only one (ablation) | L2Sep · Bin. Pack., Pack., Indep. Set, Fac. Loc. | relative time improvement: 40.4% -> 42.3%; 23.8% -> 28.5%; 70.2% -> 72.4%; 20.1% -> 29.4% | Improved |
| add a third configuration update rather than use k=2 (ablation) | L2Sep · Bin. Pack., Pack., Indep. Set, Fac. Loc. | relative time improvement: 42.3% -> 44.2%; 28.5% -> 27.8%; 72.4% -> 69.7%; 29.4% -> 29.7% | Mixed |
| replace neural UCB with offline supervised regression using four times as much data (ablation) | L2Sep · Bin. Pack., Pack., Indep. Set, Fac. Loc. | relative time improvement: 42.3% -> 40.2%; 28.5% -> 24.0%; 72.4% -> 68.7%; 29.4% -> 31.0% | Mixed |
| replace neural UCB with ε-greedy exploration (ablation) | L2Sep · Bin. Pack., Pack., Indep. Set, Fac. Loc. | relative time improvement: 42.3% -> 36.3%; 28.5% -> 25.1%; 72.4% -> 64.1%; 29.4% -> 28.1% | Hurt |
Closest papers by meaning
- Learning to Remove Cuts in Integer Linear Programming (ICML 2024 · similarity 0.88)
- Learning to Stop Cut Generation for Efficient Mixed-Integer Linear Programming (AAAI 2024 · similarity 0.87)
- MILP-StuDio: MILP Instance Generation via Block Structure Decomposition (NeurIPS 2024 · similarity 0.87)
- Learning Cut Generating Functions for Integer Programming (NeurIPS 2024 · similarity 0.87)
- Learn2Aggregate: Supervised Generation of Chvatal-Gomory Cuts Using Graph Neural Networks (AAAI 2025 · similarity 0.86)
- Contrastive Predict-and-Search for Mixed Integer Linear Programs (ICML 2024 · similarity 0.86)
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.
