Deep Reinforcement Learning for the Interval Job Shop: Checkpoints, Records and Code

Companion data and code for the article "Deep Reinforcement Learning for the Interval Job Shop Scheduling Problem: A Comparison with Genetic Programming Hyper-Heuristics across Inference Budgets". The deposit contains the final checkpoints of every training run of the article's arms (main policy at three training budgets, interval ablations, self-attention variant and width-penalizing arms); the primary CSV/JSON result files behind every table and figure, accepted and rejected records alike; the training and evaluation package (jobshop_rl) with the interval scheduling environment, the Deep Sets policy and the PPO trainer; and the verification script that recomputes every number the article prints from the primary files in this deposit. The article's supplementary material (the automatic configuration campaigns, the self-attention variant and the per-instance results) is included as a PDF. The benchmark instances and the thirty evolved GP rules the article compares against are published separately at doi:10.5281/zenodo.21716972 and are not duplicated here.