Research map / Reinforcement learning / MDPs and planning
Learning to Branch with Tree MDPs
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 |
|---|---|---|---|
| tree policy gradient with DFS instead of temporal policy gradient | REINFORCE GNN brancher · Combinatorial auctions test | final B&B tree size: 86.7 -> 86.1 | Helped |
| tree policy gradient with DFS instead of temporal policy gradient | REINFORCE GNN brancher · Set covering test | final B&B tree size: 196.3 -> 190.8 | Helped |
| tree policy gradient with objective limit instead of temporal policy gradient | REINFORCE GNN brancher · Facility location test | final B&B tree size: 393.2 -> 325.4 | Helped |
| tree policy gradient with objective limit instead of temporal policy gradient | REINFORCE GNN brancher · Facility location transfer | final B&B tree size: 679.4 -> 496.0 | Helped |
| tree policy gradient with DFS instead of temporal policy gradient | REINFORCE GNN brancher · Multiple knapsack test | final B&B tree size: 143.4 -> 135.8 | Helped |
| tree policy gradient with DFS instead of temporal policy gradient | REINFORCE GNN brancher · Maximum independent set transfer | final B&B tree size: 2768.5 -> 2970.0 | Hurt |
| tree policy gradient with objective limit instead of temporal policy gradient | REINFORCE GNN brancher · Combinatorial auctions test | final B&B tree size: 86.7 -> 87.0 | No clear effect |
| RL from scratch instead of strong branching imitation | GNN brancher · Multiple knapsack test | final B&B tree size: 228.0 -> 135.8 | Helped |
| RL from scratch instead of strong branching imitation | GNN brancher · Combinatorial auctions test | final B&B tree size: 52.2 -> 86.1 | Hurt |
| RL from scratch instead of strong branching imitation | GNN brancher · Set covering test | final B&B tree size: 51.8 -> 190.8 | Hurt |
Closest papers by meaning
- Parameterizing Branch-and-Bound Search Trees to Learn Branching Policies (AAAI 2021 · similarity 0.93)
- Learning Branching Policies for MILPs with Proximal Policy Optimization (AAAI 2026 · similarity 0.92)
- CAMBranch: Contrastive Learning with Augmented MILPs for Branching (ICLR 2024 · similarity 0.89)
- SPOT: Scalable Policy Optimization with Trees for Markov Decision Processes (NeurIPS 2025 · similarity 0.89)
- Generalization Guarantees for Learning Score-Based Branch-and-Cut Policies in Integer Programming (NeurIPS 2025 · similarity 0.88)
- Search Strategy Generation for Branch and Bound Using Genetic Programming (AAAI 2025 · 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.