Research map / Reinforcement learning / Policy gradient and actor-critic
Reinforcement Learning with Combinatorial Actions: An Application to Vehicle Routing
Arthur Delarue, Ross Anderson, Christian Tjandraatmadja
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 RLCA-16 (MIP action selection, 16-hidden-unit value net) instead of greedy | CVRP · random CVRP n=51 | mean total distance: 13.55 -> 10.65 | Improved |
| use RLCA-16 instead of Nazari et al. | CVRP · random CVRP n=21 | mean total distance: 6.40 -> 6.16 | Improved |
| RLCA-16 versus Kool et al. | CVRP · random CVRP n=51 | mean total distance: 10.62 -> 10.65 | No clear effect |
| RLCA-16 versus OR-Tools | CVRP · random CVRP n=51 | mean total distance: 10.47 -> 10.65 | Hurt |
| retain data across policy iterations (retention factor gamma near 1) (ablation) | RLCA · CVRP n=11 and n=21 | gap to optimal | Improved |
| increase sample paths per iteration (ablation) | RLCA · CVRP n=11 and n=21 | gap to optimal | Improved |
| add hidden ReLU nodes (4 vs linear model) (ablation) | value function network · CVRP n=11 and n=21 | gap to optimal | Improved |
| add mild LASSO regularization (ablation) | value function network · CVRP n=11 and n=21 | gap to optimal | No clear effect |
| include combinatorial lower bounds in objective and training (ablation) | RLCA · CVRP n=11 and n=21 | convergence of gap | Improved |
| increase value network from 0 to 16 hidden nodes (ablation) | RLCA action selection MIP (Gurobi) · CVRP n=51 | MIP runtime (s): 0.097 -> 39.3 | Hurt |
Closest papers by meaning
- Reinforcement Learning for Solving the Vehicle Routing Problem (NeurIPS 2018 · similarity 0.91)
- Preference Optimization for Combinatorial Optimization Problems (ICML 2025 · similarity 0.90)
- Reinforcement learning with combinatorial actions for coupled restless bandits (ICLR 2025 · similarity 0.88)
- Structured Reinforcement Learning for Combinatorial Decision-Making (NeurIPS 2025 · similarity 0.87)
- Improving Generalization of Neural Combinatorial Optimization for Vehicle Routing Problems via Test-Time Projection Learning (NeurIPS 2025 · similarity 0.86)
- Stochastic Q-learning for Large Discrete Action Spaces (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.
