Research map / Learning theory and algorithms / Prediction and decision losses
Deep Learning for Algorithm Portfolios
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 |
|---|---|---|---|
| Use CNN solver selection on instance images instead of the best single solver | CNN portfolio · Random SAT | % solved instances: 77.4 -> 90.7 | Helped |
| Use CNN solver selection instead of best single solver | CNN portfolio · CSP | % solved instances: 65.4 -> 82.1 | Helped |
| Use CNN solver selection instead of best single solver | CNN portfolio · Industrial SAT | % solved instances: 92.6 -> 93.0 | No clear effect |
| Use CNN solver selection instead of best single solver | CNN portfolio · Crafted SAT | % solved instances: 77.2 -> 78.2 | No clear effect |
| Use CNN selection instead of best single solver | CNN portfolio · CSP | average runtime (s): 1,501 -> 914 | Helped |
| Use CNN selection instead of best single solver | CNN portfolio · Random SAT | average runtime (s): 1,775 -> 1,143 | Helped |
| Use CNN output vector as features for CSHC instead of CNN direct selection | CNN portfolio · CSP | % solved instances: 82.1 -> 78.2 | Hurt |
| Use CNN output vector as features for CSHC instead of CNN direct selection | CNN portfolio · Industrial/Crafted SAT | % solved instances: 93.0 -> 93.7 (Industrial); 78.2 -> 79.9 (Crafted); 90.7 -> 90.1 (Random) | Mixed |
| Replace random solver choice with CNN selection (ablation) | Algorithm portfolio · Random SAT | % solved instances: 31.6 -> 90.7 | Helped |
| Shuffle clauses and sort variables lexicographically in SAT inputs (ablation) | CNN portfolio · SAT | % solved instances | No clear effect |
| Append CNN-derived features to hand-crafted features for CSHC (ablation) | CSHC · SAT/CSP | % solved instances | No clear effect |
| Increase image size from 32x32 to 256x256 (ablation) | CNN portfolio · Random SAT | % solved instances | Helped |
Closest papers by meaning
- Neural Solver Selection for Combinatorial Optimization (ICML 2025 · similarity 0.84)
- Sample Complexity of Algorithm Selection Using Neural Networks and Its Applications to Branch-and-Cut (NeurIPS 2024 · similarity 0.83)
- Solver-Free Decision-Focused Learning for Linear Optimization Problems (NeurIPS 2025 · similarity 0.82)
- UDC: A Unified Neural Divide-and-Conquer Framework for Large-Scale Combinatorial Optimization Problems (NeurIPS 2024 · similarity 0.82)
- NeuralSolver: Learning Algorithms For Consistent and Efficient Extrapolation Across General Tasks (NeurIPS 2024 · similarity 0.82)
- AutoOpt: A Dataset and a Unified Framework for Automating Optimization Problem Solving (NeurIPS 2025 · similarity 0.81)
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.