Low-Cost Labels, Reliable Choices: Rollout-Calibrated Hyper-Heuristics for Job Shop Scheduling
The article discusses a new approach to job shop scheduling using rollout-calibrated hyper-heuristics. This method focuses on optimizing label generation while ensuring the reliability of dispatching rules. The proposed selector demonstrates significant improvements in performance compared to traditional methods.
- ▪The study addresses the computational cost associated with label generation in job shop scheduling.
- ▪A gated selector is introduced that only acts when predicted improvements exceed a certain margin.
- ▪The proposed method achieves the lowest mean relative percentage deviation among learned selectors.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.23957 (cs) [Submitted on 11 May 2026] Title:Low-Cost Labels, Reliable Choices: Rollout-Calibrated Hyper-Heuristics for Job Shop Scheduling Authors:Junhao Wei, Yanxiao Li, Yifu Zhao, Zhenhong Peng, Baili Lu, Dexing Yao, Haochen Li, Qinbin He, Sio-Kei Im, Yapeng Wang, Xu Yang View a PDF of the paper titled Low-Cost Labels, Reliable Choices: Rollout-Calibrated Hyper-Heuristics for Job Shop Scheduling, by Junhao Wei and 10 other authors View PDF HTML (experimental) Abstract:Learning-assisted hyper-heuristics can select among dispatching rules while preserving the feasibility and interpretability of constructive Job Shop Scheduling Problem (JSSP) heuristics.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
