Research map / Learning theory and algorithms / Combinatorial and search optimization
SurCo: Learning Linear SURrogates for COmbinatorial Nonlinear Optimization Problems
Aaron M Ferber, Taoan Huang, Daochen Zha, Martin Schubert, Benoit Steiner, Bistra Dilkina, Yuandong Tian
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 SurCo-zero instead of the domain heuristic | Nonlinear route planning · Loose deadline | on-time success rate: 4.5% times higher success rates | Improved |
| use SurCo-zero instead of the domain heuristic | Nonlinear route planning · Normal deadline | on-time success rate: 6.5% times higher success rates | Improved |
| use SurCo-zero instead of the domain heuristic | Nonlinear route planning · Tight deadline | on-time success rate: 8.5% times higher success rates | Improved |
| use SurCo-zero instead of the domain heuristic | Nonlinear route planning · Nonlinear route planning | on-time success rate | Improved |
| use SurCo-zero instead of the domain heuristic | Embedding table sharding · Embedding table sharding | latency | Improved |
| use SurCo-prior instead of SurCo-zero | Embedding table sharding · Smaller instances (T≤40) | latency | Improved |
| use SurCo-zero instead of SurCo-prior | Embedding table sharding · Larger instances | latency | Improved |
| add instance-specific fine-tuning to the SurCo-prior initialization | SurCo-hybrid · Embedding table sharding | latency | Improved |
| use SurCo-zero instead of Pass-Through | Inverse photonic design · Inverse photonic design | valid solution rate | Improved |
| use SurCo-prior instead of Pass-Through | Inverse photonic design · Inverse photonic design | runtime: two orders of magnitude less time | Improved |
| add instance-specific fine-tuning to the SurCo-prior initialization | SurCo-hybrid · Inverse photonic design | valid solution rate | Improved |
| use SurCo-zero instead of the domain heuristic | Nonlinear route planning · Nonlinear route planning | runtime | No clear effect |
Closest papers by meaning
- Landscape Surrogate: Learning Decision Losses for Mathematical Optimization Under Partial Information (NeurIPS 2023 · similarity 0.88)
- Unsupervised Learning for Combinatorial Optimization with Principled Objective Relaxation (NeurIPS 2022 · similarity 0.87)
- Learning Surrogates for Offline Black-Box Optimization via Gradient Matching (ICML 2024 · similarity 0.85)
- Geometric Algorithms for Neural Combinatorial Optimization with Constraints (NeurIPS 2025 · similarity 0.84)
- Solving hidden monotone variational inequalities with surrogate losses (ICLR 2025 · similarity 0.84)
- Accelerating Cutting-Plane Algorithms via Reinforcement Learning Surrogates (AAAI 2024 · similarity 0.84)
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.
