Research map / Learning theory and algorithms / Prediction and decision losses
ATLAS: Alibaba Dataset and Benchmark for Learning-Augmented Scheduling
Zhiyun Jiang, Tianming Zhao, Chunqiu xia, Albert Zomaya
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 |
|---|---|---|---|
| add task-signature features and empirical Bayes shrinkage via LGBM meta-stacking | Single-Stage baseline · ATLAS PAI prediction | Cov@50 (%): 36.2 -> 70.6 | Improved |
| use classification-first routing (Two-Stage) for recurring jobs | LGBM-Meta · recurring jobs | Cov@50 (%): 77.5 -> 78.0 | Improved |
| use signature-based predictions instead of HRAS history | HRAS · ATLAS PAI prediction | RMSLE: 1.373 -> 0.871 | Improved |
| use Weighted-Rec temporal ensemble instead of meta-stacking | LGBM-Meta · ATLAS PAI prediction | Cov@25 (%): 51.2 -> 39.9 | Hurt |
| fully commit to predicted ranking (SPJF) rather than hedge with predicted round-robin (PRR) | Meta predictor · Total completion time | ratio to SRPT: 1.292 -> 1.106 | Improved |
| replace HRAS predictions with Meta predictions in SPJF | HRAS-SPJF · Total completion time | ratio: 1.686 -> 1.106 | Improved |
| use predicted-size SPRPT instead of non-preemptive LAS/FB | LAS/FB · Max-stretch | rho_S,max: 205.28 -> 18.13 | Improved |
| use SPRPT instead of EDF-P with Meta predictions | Meta-EDF · Max-stretch | rho_S,max: 204.66 -> 18.13 | Improved |
| use EDF-P with Meta predictions instead of LAS/FB for median stretch | LAS/FB · Median stretch | rho_med: 187.31 -> 4.504 | Improved |
| use HRAS predictions with EDF-P instead of Meta predictions | Meta-EDF · Max-stretch | rho_S,max: 204.66 -> 1700.5 | Hurt |
| use Iso predictor for LPPT instead of Meta | Meta-LPPT · Makespan | ratio: 1.689 -> 1.574 | Improved |
Closest papers by meaning
- Scheduling of Time-Varying Workloads Using Reinforcement Learning (AAAI 2021 · similarity 0.85)
- Algorithms with Calibrated Machine Learning Predictions (ICML 2025 · similarity 0.84)
- Non-Clairvoyant Scheduling with Progress Bars (NeurIPS 2025 · similarity 0.84)
- DynaSchedBench: Calibrated Dynamic Scheduling Benchmarks and Observability Paradox in LLM-based Scheduling Agents (ICML 2026 · similarity 0.84)
- Non-clairvoyant Scheduling with Partial Predictions (ICML 2024 · similarity 0.83)
- SkipPredict: When to Invest in Predictions for Scheduling (NeurIPS 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.
