Research map / Learning theory and algorithms / Combinatorial and search optimization
Deep Learning for Cost-Optimal Planning: Task-Dependent Planner Selection
Silvan Sievers, Michael Katz, Shirin Sohrabi, Horst Samulowitz, Patrick Ferber
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 discrete runtime classes instead of binary solved target (ablation) | planner selection CNN · IPC 2018 cost-optimal test (240 tasks) | pairwise settings won (of 12): 10 of 12 settings | Improved |
| use binary target instead of other prediction methods (ablation) | planner selection CNN · IPC 2018 cost-optimal test (240 tasks) | settings won | Hurt |
| extend planner collection from 17 (CD) to all 29 (CA) (ablation) | planner selection CNN · IPC 2018 cost-optimal test (240 tasks) | settings won (of 16): CD beats CA in 12 of 16 | Hurt |
| use minimal 9-planner set (CC) instead of CA (ablation) | planner selection CNN · IPC 2018 cost-optimal test (240 tasks) | settings won (of 16): CC beats CA in 10 of 16 | Improved |
| domain-preserving split instead of random split for validation (ablation) | planner selection CNN · IPC 2018 cost-optimal test (240 tasks) | settings won | No clear effect |
| random split with validation instead of domain-preserving split with validation (ablation) | planner selection CNN · IPC 2018 cost-optimal test (240 tasks) | settings won: 7 of 12 vs 5 | Improved |
| use validation set instead of no validation (domain split) (ablation) | planner selection CNN · IPC 2018 cost-optimal test (240 tasks) | settings won: 5 vs 7 | No clear effect |
| learned CNN selector (CD) vs random planner from CD | planner selection CNN · IPC 2018 cost-optimal test (240 tasks) | coverage (%): 42.8 -> 50.0 | Improved |
| learned CNN selector (CC) vs random planner from CC | planner selection CNN · IPC 2018 cost-optimal test (240 tasks) | coverage (%): 50.3 -> 49.5 to 58.3 | Mixed |
| best setting vs Delfi1 (winner IPC 2018) | planner selection CNN · IPC 2018 cost-optimal test (240 tasks) | coverage (%): 60.0 -> 58.3 | Hurt |
Closest papers by meaning
- Graph Learning for Numeric Planning (NeurIPS 2024 · similarity 0.84)
- State Encodings for GNN-Based Lifted Planners (AAAI 2025 · similarity 0.83)
- Task-Agnostic Pre-training and Task-Guided Fine-tuning for Versatile Diffusion Planner (ICML 2025 · similarity 0.83)
- Planning Anything with Rigor: General-Purpose Zero-Shot Planning with LLM-based Formalized Programming (ICLR 2025 · similarity 0.83)
- Visual Planning: Let's Think Only with Images (ICLR 2026 · similarity 0.83)
- Stop! Planner Time: Metareasoning for Probabilistic Planning Using Learned Performance Profiles (AAAI 2024 · similarity 0.83)
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.
