Research map / Reinforcement learning / Policy gradient and actor-critic
Learning Large Neighborhood Search Policy for Integer Programming
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 |
|---|---|---|---|
| learned LNS policy vs SCIP at same 200s limit | LNS with SCIP repair · MC | objective: -852.57 -> -889.61 | Helped |
| learned destroy policy vs R-LNS hand-crafted rule | LNS with SCIP repair · CA | objective: -109550 -> -112666 | Helped |
| learned policy vs FT-LNS imitation learning | LNS with SCIP repair · SC | primal gap (%): 3.02 -> 0.68 | Helped |
| train on small instances, test on 4x larger MIS | learned LNS policy · MIS_4 (200s) | objective: -2652.56 -> -2735.86 | Helped |
| use learned policy with Gurobi repair vs Gurobi alone | Gurobi · MC | objective: -863.91 -> -888.97 | Helped |
| SCIP given 1000s vs learned LNS at 200s | SC · SC | objective: 550.68 vs 551.50 | Hurt |
| fully-shared network instead of semi-shared policy network (ablation) | learned LNS policy · SC, CA, MC, MIS training curves | validation objective | Mixed |
| use FT-LNS imitation policy on larger instances (SC_2, Gurobi) | FT-LNS · SC_2 Gurobi | primal gap (%): 2.43 -> 17.99 | Hurt |
Closest papers by meaning
- Reinforcement Learning for Integer Programming: Learning to Cut (ICML 2020 · similarity 0.89)
- BTBS-LNS: Binarized-Tightening, Branch and Search on Learning LNS Policies for MIP (ICLR 2025 · similarity 0.89)
- A General Large Neighborhood Search Framework for Solving Integer Linear Programs (NeurIPS 2020 · similarity 0.88)
- Searching Large Neighborhoods for Integer Linear Programs with Contrastive Learning (ICML 2023 · similarity 0.87)
- Large Language Model-driven Large Neighborhood Search for Large-Scale MILP Problems (ICML 2025 · similarity 0.86)
- Regularized Large Neighborhood Search (NeurIPS 2026 · similarity 0.85)
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.