Research map / Learning theory and algorithms / Combinatorial and search optimization
Learning to Solve Orienteering Problem with Time Windows and Variable Profits
Songqun Gao, Zanxi Ruan, Patrick Floor, Marco Roveri, Luigi Palopoli, Daniele Fontanelli
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 |
|---|---|---|---|
| add spatial encoder (SE) (ablation) | POMO baseline · OPTWVP n=50 TW=100 | optimality gap (%): 25.3% -> 23.0% | Improved |
| add spatial encoder (SE) (ablation) | POMO baseline · OPTWVP n=50 TW=500 | optimality gap (%): 39.6% -> 31.2% | Improved |
| add service time optimization (STO) (ablation) | Baseline + SE · OPTWVP n=50 TW=100 | optimality gap (%): 23.0% -> 2.28% | Improved |
| add service time optimization (STO) (ablation) | Baseline + SE · OPTWVP n=50 TW=500 | optimality gap (%): 31.2% -> 1.88% | Improved |
| add supervised loss (SL) with pTAR (ablation) | Baseline + SE + STO · OPTWVP n=50 TW=100 | optimality gap (%): 2.28% -> 1.06% | Improved |
| add supervised loss (SL) with pTAR (ablation) | Baseline + SE + STO · OPTWVP n=50 TW=500 | optimality gap (%): 1.88% -> 0.83% | Improved |
| add STO module to GFACS | GFACS · OPTWVP n=100 TW=100 | optimality gap (%): 13.4% -> 3.38% | Improved |
| extend ILS search time from 10s to 50s (ablation) | ILS · OPTWVP n=100 TW=100 | score: 24.9 -> 25.0 | No clear effect |
| extend ILS search time from 10s to 200s (ablation) | ILS · OPTWVP n=50 TW=500 | score: 47.8 -> 47.8 | No clear effect |
| extend ILS search time from 10s to 200s (ablation) | ILS · OPTWVP n=100 TW=500 | score: 85.0 -> 83.7 | Hurt |
| vary cost function parameters beta1, beta2 over 100-5000 (ablation) | DeCoST · OPTWVP n=50 TW=100 | optimality gap (%): 0.81% - 0.97% | No clear effect |
| use fixed pre-reserved service time ratio (70%) instead of learned STD and STO (ablation) | two-stage decomposition · OPTWVP mixed test sets | optimality gap (%): 3.87% | Hurt |
Closest papers by meaning
- DPN: Decoupling Partition and Navigation for Neural Solvers of Min-max Vehicle Routing Problems (ICML 2024 · similarity 0.81)
- Learning-Augmented Algorithms for the Bahncard Problem (NeurIPS 2024 · similarity 0.81)
- Continual Optimization with Symmetry Teleportation for Multi-Task Learning (NeurIPS 2025 · similarity 0.81)
- Step-by-Step Optimization-like Reasoning in LLMs over Expanding Search Spaces (NeurIPS 2026 · similarity 0.81)
- Learning-Guided Rolling Horizon Optimization for Long-Horizon Flexible Job-Shop Scheduling (ICLR 2025 · similarity 0.80)
- Free from Bellman Completeness: Trajectory Stitching via Model-based Return-conditioned Supervised Learning (ICLR 2024 · similarity 0.80)
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.
