Research map / Reinforcement learning / Visual and representation RL
Factorized Scheduling Principle: Learning Interpretable and Transferable Policies via Structured Additive Functions
Hong Je-Gal, Hyun-Suk Lee
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 |
|---|---|---|---|
| transfer FSP trained at N=10 zero-shot to N=20 instead of using Whittle index | FSP · Wireless scheduling | average reward: -0.888 -> -0.781 | Improved |
| use FSP instead of Whittle index at small N | FSP · Wireless scheduling, N=5 | average reward: 0.096 -> 0.085 | Hurt |
| use FSP instead of Whittle index | FSP · Warehouse scheduling, N=20 | average reward: 0.605 -> 0.649 | Improved |
| use FSP instead of Whittle index | FSP · Inventory scheduling, N=20 | average reward: -0.921 -> -0.726 | Improved |
| use FSP instead of Whittle index at training size | FSP · Inventory scheduling, N=10 (ID) | average reward: -0.344 -> -0.353 | Hurt |
| use zero-shot FSP instead of per-size trained DQN | FSP · Warehouse scheduling, N=10 (ID) | average reward: 0.883 -> 0.823 | Mixed |
| use FSP instead of per-size trained NAM | FSP · Warehouse scheduling, N=5 | average reward: -9.272 -> 0.630 | Improved |
| use per-size trained QR-DQN instead of zero-shot FSP | QR-DQN · Inventory scheduling, N=5 | average reward: -0.206 -> -0.127 | Mixed |
| increase candidate count from 4 to 32 for FSP trained at N=8 | FSP · Synthetic transfer | ranking consistency: 0.9940 -> 0.9941 | No clear effect |
| increase candidate count from 4 to 32 for FSP trained at N=8 | FSP · Synthetic transfer | reward gap to oracle: 7.9x10^-5 -> 6.15x10^-4 | Mixed |
Closest papers by meaning
- Towards Generalizable Multi-Policy Optimization with Self-Evolution for Job Scheduling (NeurIPS 2025 · similarity 0.84)
- Inductive-bias-driven Reinforcement Learning For Efficient Schedules in Heterogeneous Clusters (ICML 2020 · similarity 0.84)
- Q-MMR: Off-Policy Evaluation via Recursive Reweighting and Moment Matching (NeurIPS 2026 · similarity 0.83)
- Learning-Augmented Priority Queues (NeurIPS 2024 · similarity 0.82)
- DynaSchedBench: Calibrated Dynamic Scheduling Benchmarks and Observability Paradox in LLM-based Scheduling Agents (ICML 2026 · similarity 0.82)
- SPOT: Scalable Policy Optimization with Trees for Markov Decision Processes (NeurIPS 2025 · similarity 0.82)
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.
