Research map / Learning theory and algorithms / Combinatorial and search optimization
Simulation-guided Beam Search for Neural Combinatorial Optimization
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 |
|---|---|---|---|
| add SGBS+EAS hybrid instead of EAS alone | POMO · CVRP n=100 | gap to HGS: 0.23% -> 0.11% | Helped |
| add SGBS+EAS hybrid instead of EAS alone | POMO · TSP n=100 | optimality gap: 0.044% -> 0.024% | Helped |
| use SGBS (4,4) instead of sampling at equal candidate budget | POMO · CVRP n=200 | gap to HGS: 6.48% -> 3.68% | Helped |
| use SGBS (4,4) instead of sampling | POMO · CVRP n=100 | gap to HGS: 0.64% -> 0.62% | No clear effect |
| use beam search or sampling instead of greedy under distribution shift | POMO · CVRP n=200 | gap to HGS: 5.19% -> 6.48% | Hurt |
| add SGBS to EAS on FFSP | MatNet · FFSP20 | gap: 0.30 -> 0.22 | Helped |
| add SGBS to EAS on FFSP100 | MatNet · FFSP100 | gap: 0.71 -> 0.33 | Helped |
| select top beta*gamma children by accumulated NLP-style probability instead of top gamma per node (ablation) | SGBS · CVRP | search performance | Hurt |
| over-train pre-trained policy before SGBS+EAS (ablation) | SGBS+EAS · CVRP n=100 | solution quality | Hurt |
| increase beam width and expansion factor of SGBS (ablation) | SGBS · routing | solution quality | No clear effect |
Closest papers by meaning
- Probing Neural Combinatorial Optimization Models (NeurIPS 2025 · similarity 0.88)
- Interpreting Neural Combinatorial Optimization via Evolving Programmatic Bottlenecks (NeurIPS 2026 · similarity 0.87)
- Neural Solver Selection for Combinatorial Optimization (ICML 2025 · similarity 0.87)
- PolyNet: Learning Diverse Solution Strategies for Neural Combinatorial Optimization (ICLR 2025 · similarity 0.87)
- Fast T2T: Optimization Consistency Speeds Up Diffusion-Based Training-to-Testing Solving for Combinatorial Optimization (NeurIPS 2024 · similarity 0.87)
- Scaling Combinatorial Optimization Neural Improvement Heuristics with Online Search and Adaptation (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.