Research map / Reinforcement learning / Visual and representation RL
Learning to Search Feasible and Infeasible Regions of Routing Problems with Flexible Neural k-Opt
Yining Ma, Zhiguang Cao, Yeow Meng Chee
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 flexible k-opt exchanges with the RDS decoder rather than the DACT-U 2-opt design | NeuOpt · TSP-100 | Gap: 0.32% -> 0.24% | Improved |
| Use GIRE instead of feasibility masking | NeuOpt · CVRP-20 | Gap: 0.27% -> 0.05% | Improved |
| Increase the maximum basis-move limit from K=2 to K=4 (ablation) | NeuOpt-GIRE · CVRP-20 | Objective: 6.167 -> 6.163 | Improved |
| Remove GRUs from the RDS decoder (ablation) | NeuOpt · TSP-100 | Objective: 7.798 -> 7.804 | Hurt |
| Remove GRUs from the RDS decoder (ablation) | NeuOpt-GIRE · CVRP-20 | Objective: 6.163 -> 6.165 | Hurt |
| Remove move stream μ (ablation) | NeuOpt · TSP-100 | Objective: 7.798 -> 7.806 | Hurt |
| Remove move stream μ (ablation) | NeuOpt-GIRE · CVRP-20 | Objective: 6.163 -> 6.165 | Hurt |
| Remove edge stream λ (ablation) | NeuOpt · TSP-100 | Objective: 7.798 -> 7.799 | Hurt |
| Remove edge stream λ (ablation) | NeuOpt-GIRE · CVRP-20 | Objective: 6.163 -> 6.164 | Hurt |
| Add dynamic data augmentation during inference (ablation) | NeuOpt · TSP-100 | Gap: 0.09% -> 0.05% | Improved |
| Add dynamic data augmentation during inference (ablation) | NeuOpt-GIRE · CVRP-100 | Gap: 1.00% -> 0.87% | Improved |
| Add dynamic data augmentation during inference (ablation) | NeuOpt · TSP-100 | Gap: 0.04% -> 0.02% | Improved |
Closest papers by meaning
- Memory-Enhanced Neural Solvers for Routing Problems (NeurIPS 2025 · similarity 0.91)
- Learning to Handle Complex Constraints for Vehicle Routing Problems (NeurIPS 2024 · similarity 0.90)
- LINC: Decoupling Local Consequence Scoring from Hidden Matching in Constructive Neural Routing (NeurIPS 2026 · similarity 0.89)
- MTL-KD: Multi-Task Learning Via Knowledge Distillation for Generalizable Neural Vehicle Routing Solver (NeurIPS 2025 · similarity 0.89)
- Improving Generalization of Neural Combinatorial Optimization for Vehicle Routing Problems via Test-Time Projection Learning (NeurIPS 2025 · similarity 0.89)
- Neural Solver Selection for Combinatorial Optimization (ICML 2025 · similarity 0.88)
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.
