Pith. sign in

REVIEW 5 major objections 5 minor 63 references

An Expectation-Maximization Algorithm-based Autoregressive Model for the Fuzzy Job Shop Scheduling Problem

T0 review · 5 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read An autoregressive model trained by an EM-style self-labeling loop can solve the fuzzy job shop scheduling problem, producing better fuzzy makespan than four evolutionary heuristics while running at least three times faster.

desk verdict A plausible first end-to-end learned solver for fuzzy job shop scheduling, undercut by an undefined fuzzy makespan and a sloppy table entry. read the letter →

arxiv 2502.00018 v1 pith:LMKNAF2F submitted 2025-01-11 cs.AI

Yijian Wang , Tongxian Guo , Zhaoqiang Liu This is my paper · ORCID
classification cs.AI
keywords fuzzyjobshopschedulingautoregressivemodelexpectation-maximizationneuralcombinatorialoptimizationdisjunctivegraphtriangularnumbersself-labelingmakespanminimization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that a neural network can be the primary solver, rather than a helper, for the fuzzy job shop scheduling problem, where processing times are triangular fuzzy numbers. The authors model the FJSSP as a generative task: an autoregressive network with a disjunctive-graph encoder outputs a job sequence, and training needs no optimal schedules as labels. Instead, an expectation-maximization loop alternates an E-step that samples candidate schedules from the current model and keeps the one with the smallest fuzzy makespan, and an M-step that updates the model to imitate that chosen schedule. Across the benchmark instances they report, the trained model achieves lower fuzzy makespan than four established fuzzy-scheduling algorithms and runs at least three times faster in testing. If the comparison is right, learned one-shot generation is a practical alternative to population-based search for scheduling under uncertainty.

What carries the argument

The load-bearing mechanism is the EM-style self-labeling loop around an autoregressive generative model. At each epoch, the current network samples $K$ candidate schedules per instance; the candidate with the smallest fuzzy makespan becomes the pseudo-ground-truth, and the network is updated by maximum likelihood on those pseudo-labels. The network itself combines a hand-crafted 18-dimensional fuzzy prior per operation, a two-layer graph attention network over the disjunctive graph, a multi-head-attention state network, and a feedforward decision network that outputs a softmax over jobs. Fuzzy numbers are ranked by $Z(\tilde{A}) = V_\beta(\tilde{A}) + \omega S(\tilde{A})$ with $\beta=0.5$ and $\omega=0.4$, the rule used to compare fuzzy makespan values when selecting pseudo-labels.

What would settle it

Recompute every reported $FMS(\pi)$ with an explicit fuzzy arithmetic: propagate $\tilde{s}_i+\tilde{t}_i$ by triangular addition, define makespan via pairwise max under the paper's $Z(\tilde{A}) = V_\beta(\tilde{A})+\omega S(\tilde{A})$ ranking, and check that each reported value is a valid triangular fuzzy number with $a_1 \le a_2 \le a_3$. The FT20 row reports $(1129,1223,1141)$, which violates this shape; if that survives in a re-run, or if any column winner changes under the paper's own ranking rule, the claim that EMARM outperforms all comparators is not established.

Watch

Extended reading notes

Core claim

The central claim is that an autoregressive model $p_\theta(\pi|I)$ trained by an EM-style self-labeling loop can solve the fuzzy job shop scheduling problem end-to-end. The paper's EMARM represents an instance as a disjunctive graph, feeds each operation a hand-crafted 18-dimensional fuzzy prior, encodes it with a two-layer graph attention network, and decodes a schedule by sequential job selection. Because the true optimal schedule $\pi^*$ is unknown, the E-step samples $K$ schedules from the current policy and keeps the candidate minimizing the fuzzy makespan $FMS(\pi)$, and the M-step maximizes $\log p_\theta(\hat\pi|I)$ for these pseudo-labels. The paper reports that this yields better fuzzy makespan than NSODE, HADE, PSO, and GA on its benchmark set, with a threefold or larger testing-time reduction, and interprets the result as evidence that fuzzy scheduling problems can be handled by learned generative models rather than by evolutionary search.

Load-bearing premise

The load-bearing premise is that $FMS(\pi)$, the fuzzy makespan used to choose pseudo-labels and to rank competitors, is a faithful and consistently computed version of the stated fuzzy objective, yet the paper never defines it and handles subtraction and division by ad hoc defuzzification.

Editorial extensions

If this is right

  • After training, EMARM solves new instances by forward passes only, with no iterative search; the reported runtime advantage is a direct corollary.
  • The method is claimed to generalize to benchmark sizes beyond the training distribution, since 30x20 test instances appear although training sizes go up to 20x20.
  • The EM-style self-labeling loop gives a general route for training neural combinatorial optimization models on any problem whose objective can be evaluated but whose optimal solutions are unknown.
  • Because the model is generative, it can sample multiple candidate schedules at test time and re-rank them by fuzzy makespan, which the paper does by increasing the sampling count from 256 to 512.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Inference: the reported quality comparison depends on an FMS function that the paper never defines; re-running the tables under alternative fuzzy ranking rules (integral value, centroid, or the stated Z-value) would show whether the advantage is an artifact of the chosen fuzzy arithmetic.
  • Inference: the self-labeling loop is a form of reward-weighted policy iteration, so its behavior is sensitive to K and to the quality of the initial random policy; a useful follow-up is to plot solution quality against K and against training epoch to see whether the loop reliably escapes mediocre policies.
  • Inference: the malformed FT20 entry suggests the fuzzy arithmetic was not automatically validated; a reproducible implementation that checks that every triangular fuzzy number satisfies $a_1 \le a_2 \le a_3$ would be a prerequisite for trusting the benchmark comparison.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper proposes EMARM, an autoregressive neural model for the fuzzy job shop scheduling problem (FJSSP). Processing times are triangular fuzzy numbers, and the model generates job sequences autoregressively from a disjunctive-graph representation with hand-crafted fuzzy priors. Training alternates between sampling K schedules from the current model and selecting the best according to the fuzzy makespan (E-step), then maximizing the likelihood of those pseudo-labels (M-step). Experiments on 42 benchmark instances compare EMARM against NSODE, HADE, GA, and PSO, reporting lower fuzzy makespan values in Table III and substantially lower runtime in Table II. The paper concludes that modeling FJSSP as a generative task and using the EM-style self-labeling loop yields a practical primary solver for fuzzy scheduling.

Significance. If correct, this is a meaningful step: it is among the first attempts to use a learned autoregressive model as the primary solver for the fuzzy JSSP, rather than as a helper inside a metaheuristic. The idea of treating the optimal schedule as a latent variable and training via self-labeling is attractive, especially because ground-truth optimal schedules are unavailable. The paper reports large empirical advantages in both solution quality and runtime. However, the correctness of these claims depends on a precise and consistent definition of the fuzzy makespan FMS(·), which is currently absent. The reported comparison also lacks variance information for EMARM, and one table entry is not a valid triangular fuzzy number. These issues prevent the headline claim from being independently confirmed. With a precise definition, a rerun of the benchmarks, and appropriate statistics, the approach could be a valuable contribution to NCO for fuzzy scheduling.

major comments (5)
  1. [§II.B, Eq. (10); Algorithm 1 line 7; Table III] The function FMS(·) is never defined. Section II.B states the objective as max_{i∈O}(s̃_i + t̃_i), and Section II.A defines a pairwise max operation via the ranking rule of Eq. (5), but the manuscript does not specify how this fuzzy max is computed over all operations of a schedule, how the resulting makespan TFN is formed, or how the subtraction and division operations mentioned in Section II.A are actually executed. Algorithm 1 line 7 uses FMS(π) to select the pseudo-label in the E-step, and Table III reports FMS values as the evaluation metric. An undefined objective makes the central claim of Section IV.C ("EMARM outperforms all comparative algorithms") unfalsifiable: the reader cannot reproduce the E-step selection, the test-time decoding, or the benchmark comparison. This must be fixed by giving an exact computational definition of FMS(·), including the treatment of fuzzy max and any defuzzification steps.
  2. [Table III, FT20 row] The EMARM entry for FT20 is (1129, 1223, 1141), which violates the defining condition of a triangular fuzzy number, a1 ≤ a2 ≤ a3 (here 1223 > 1141). If this is a typo, it indicates that the reported fuzzy arithmetic was not systematically verified; if it is not a typo, the value is not a valid TFN. Either way, the entry must be corrected and the entire table rechecked for similar inconsistencies. This is a concrete sign that the undefined FMS definition has real consequences for the reported results.
  3. [§IV.C, Table III] The comparison reports only point values for EMARM, despite the method being stochastic: the paper states that K is increased to 512 during testing, and the decoder samples from the learned distribution. The evolutionary baselines are averaged over 30 runs, but no variance, standard deviation, or statistical test is given for EMARM. Without this information, the claim that EMARM outperforms all comparative algorithms is not supported; the observed margins could be within the noise of a single or few runs. The authors should report the mean and standard deviation (or median and quartiles) over multiple test-time decoding runs, and preferably a paired statistical test across the 42 instances.
  4. [§II.A, Eq. (5); §IV.B] The ranking rule Z(Ã) = Vβ(Ã) + ωS(Ã) is central: it defines the fuzzy max used in FMS, the pseudo-label selection in the E-step, and the final comparison in Table III. The weights β = 0.5 and ω = 0.4 are imported from reference [11] without any sensitivity analysis. If the ranking of different schedules or algorithms changes with β and ω, the reported superiority may be an artifact of these particular weights. The authors should either justify these values for the benchmark instances or provide a sensitivity study showing that the conclusions are robust over a reasonable range of β and ω.
  5. [§III.E, Algorithm 1, Eq. (32)] The procedure is described as an EM algorithm, but the E-step in Algorithm 1 does not compute the posterior expectation over latent variables; it samples K schedules from the current model and picks the one with minimum FMS. This is a hard self-labeling / self-training heuristic, not the EM update justified by the Jensen lower bound in Eq. (32). The paper does not show that this selection maximizes the lower bound, nor does it provide a convergence or consistency argument. Since the claimed novelty of the training method is its EM interpretation, this mismatch should be addressed explicitly, either by repositioning the method as a self-training variant or by providing a theoretical justification for the hard-selection step.
minor comments (5)
  1. [§IV.C] There is a typo in the text: "fuzzy makeapsn" should be "fuzzy makespan."
  2. [§IV.B] The architectural description is internally inconsistent: it states "In GAT 1, the size of each head is set to 64 and their outputs are concatenated" (implying 3×64 = 192), yet then says "Therefore, h = 16 + 128 = 144." The relation between the head outputs and the reported dimensionality is unclear and should be reconciled with Eq. (22).
  3. [§III.B, Eqs. (16)-(17)] The indices in Eqs. (16) and (17) are ambiguous: Eq. (16) appears to sum from St_i to i in the numerator and from St_i to End_i in the denominator, but this is not stated clearly. Please clarify the intended ranges.
  4. [References] Reference [53] contains a formatting error: "10 721–10 734" should be "10721–10734."
  5. [§IV.C] The description of the experimental protocol does not state how many test-time samples are used to obtain each reported EMARM value, whether the same random seed is used across instances, or whether the 30-run average for baselines uses the same stopping criterion as the 100-iteration setting described in the text. These details should be provided for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the training loop is self-referential by design, but the central comparison is evaluated against fixed benchmark objective values.

full rationale

The paper's training loop is explicitly self-labeling: Algorithm 1 samples K schedules from the current model and selects the best according to FMS(π), then maximizes the likelihood of those pseudo-labels. This is a self-referential training procedure, but it is not a circular derivation of the paper's central claim. The pseudo-labels are selected by an external objective function evaluated on each instance and schedule, and the final evaluation in Table III compares EMARM against NSODE, HADE, PSO, and GA on fixed benchmark instances using reported fuzzy makespan values. No benchmark value is used as a training label, and no fitted parameter is renamed as a prediction. The EM lower bound in Eq. (32) is used heuristically, and the M-step ignores part of the bound, but this is a methodological approximation rather than a circular reduction. The only self-citation is reference [62] in the future-work paragraph, which is not load-bearing. The undefined FMS(·), the suspicious FT20 entry, and the loose handling of fuzzy arithmetic are serious correctness and reproducibility concerns, but they do not make the derivation circular under the criteria used here: the claimed superiority is not equivalent to the training inputs by construction, and the comparison is against independent benchmark results.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central training loop is self-referential, and the ranking weights and sampling count are used to define the objective. No new physical entities are introduced. The fuzzy prior vector is a feature construction, not an invented entity.

free parameters (3)
  • Fuzzy ranking weights beta and omega = beta=0.5, omega=0.4
    Set according to [11] in Section IV.B; define the Z-value in Eq. (5) used to order fuzzy completion times and select the best sampled schedule in Algorithm 1.
  • Number of E-step samples K = 256 (training), 512 (testing)
    Chosen in Section IV.B; larger K makes the pseudo-label selection closer to the best of K samples, but no convergence analysis ties K to the EM lower bound.
  • Training epochs T = 30
    Fixed in Section IV.B; stopping is by epoch count, not by convergence of the lower bound.
assumptions (4)
  • standard math A neural network p_theta can approximate the conditional schedule distribution in Eq. (12).
    Standard function approximation assumption behind the autoregressive model; unproved but typical in NCO.
  • domain assumption Triangular fuzzy arithmetic with componentwise addition and ranking-based max defines a valid completion-time calculus.
    Section II.A; used to compute FC and FMS, though the max operation and FMS are never fully specified.
  • ad hoc to paper The optimal schedule pi* can be treated as a latent variable, and K samples with best FMS form a valid EM E-step.
    Algorithm 1 and Section III.E; no convergence guarantee is given, and the M-step drops the p(y|x) term from Eq. (32).
  • ad hoc to paper The hand-crafted 18-dimensional prior vector in Eqs. (14)-(21) is sufficient for the encoder.
    No ablation study; the design is justified only by intuition in Section III.B.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An Expectation-Maximization Algorithm-based Autoregressive Model for the Fuzzy Job Shop Scheduling Problem." pith.science (2026). https://pith.science/paper/LMKNAF2F

@misc{pith2026250200018,
  author       = {Pith},
  title        = {Pith review of: An Expectation-Maximization Algorithm-based Autoregressive Model for the Fuzzy Job Shop Scheduling Problem},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LMKNAF2F}},
  note         = {Machine review of arXiv:2502.00018}
}
read the original abstract

The fuzzy job shop scheduling problem (FJSSP) emerges as an innovative extension to the job shop scheduling problem (JSSP), incorporating a layer of uncertainty that aligns the problem more closely with the complexities of real-world manufacturing environments. This improvement increases the computational complexity of deriving the solution while improving its applicability. In the domain of deterministic scheduling, neural combinatorial optimization (NCO) has recently demonstrated remarkable efficacy. However, its application to the realm of fuzzy scheduling has been relatively unexplored. This paper aims to bridge this gap by investigating the feasibility of employing neural networks to assimilate and process fuzzy information for the resolution of FJSSP, thereby leveraging the advancements in NCO to enhance fuzzy scheduling methodologies. To achieve this, we approach the FJSSP as a generative task and introduce an expectation-maximization algorithm-based autoregressive model (EMARM) to address it. During training, our model alternates between generating scheduling schemes from given instances (E-step) and adjusting the autoregressive model weights based on these generated schemes (M-step). This novel methodology effectively navigates around the substantial hurdle of obtaining ground-truth labels, which is a prevalent issue in NCO frameworks. In testing, the experimental results demonstrate the superior capability of EMARM in addressing the FJSSP, showcasing its effectiveness and potential for practical applications in fuzzy scheduling.

Figures

Figures reproduced from arXiv: 2502.00018 by the authors.

Figure 1
Figure 1. Disjunctive graph model. (a) illustrates a 3 × 3 FJSSP instance. The [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 51 canonical work pages

  1. [11]

    A hybrid cooperative coevolution algorithm for fuzzy flexible job shop scheduling,

    L. Sun, L. Lin, M. Gen, and H. Li, “A hybrid cooperative coevolution algorithm for fuzzy flexible job shop scheduling,” IEEE Transactions on Fuzzy Systems , vol. 27, no. 5, pp. 1008–1022, 2019

    work page 2019

  2. [1]

    A novel mathematical model for the flexible job-shop scheduling problem with limited automated guided vehicles,

    Y . Yao, Q. Liu, L. Fu, X. Li, Y . Yu, L. Gao, and W. Zhou, “A novel mathematical model for the flexible job-shop scheduling problem with limited automated guided vehicles,” IEEE Transactions on Automa- tion Science and Engineering , early, access, 2024. doi: 10.1109/TAS E.2024.3356255

    work page arXiv 2024

  3. [2]

    Knowledge-based lot-splitting optimization method for flexible job shops considering energy consumption,

    Z. Tian, X. Jiang, G. Tian, Z. Li, and W. Liu, “Knowledge-based lot-splitting optimization method for flexible job shops considering energy consumption,” IEEE Transactions on Automation Science and Engineering, vol. 21, no. 3, pp. 4864–4785, 2024

    work page 2024

  4. [3]

    Multitask multiobjective genetic programming for automated scheduling heuristic learning in dy- namic flexible job-shop scheduling,

    F. Zhang, Y . Mei, S. Nguyen, and M. Zhang, “Multitask multiobjective genetic programming for automated scheduling heuristic learning in dy- namic flexible job-shop scheduling,” IEEE Transactions on Cybernetics, vol. 53, no. 7, pp. 4473–4486, 2023

    work page 2023

  5. [4]

    Multiobjec- tive optimization of energy-efficient job-shop scheduling with dynamic reference point-based fuzzy relative entropy,

    L. He, R. Chiong, W. Li, S. Dhakal, Y . Cao, and Y . Zhang, “Multiobjec- tive optimization of energy-efficient job-shop scheduling with dynamic reference point-based fuzzy relative entropy,” IEEE Transactions on Industrial Informatics, vol. 18, no. 1, pp. 600–610, 2022

    work page 2022

  6. [5]

    Huang, S

    M. Huang, S. Huang, B. Du, J. Guo, and Y . Li, “Fuzzy superposition operation and knowledge-driven co-evolutionary algorithm for integrated production scheduling and vehicle routing problem with soft time windows and fuzzy travel times,” IEEE Transactions on Fuzzy Systems , early, access, 2024. doi: 10.1109/TFUZZ.2024.3388003

    arXiv 2024

  7. [6]

    Self-adaptive multi-objective evolutionary algorithm for flexible job shop scheduling with fuzzy processing time,

    R. Li, W. Gong, and C. Lu, “Self-adaptive multi-objective evolutionary algorithm for flexible job shop scheduling with fuzzy processing time,” Computers & Industrial Engineering , vol. 168, p. 108099, 2022

    work page 2022

  8. [7]

    Bi-population balancing multi-objective algorithm for fuzzy flexible job shop with energy and transportation,

    J. Li, Y . Han, K. Gao, X. Xiao, and P. Duan, “Bi-population balancing multi-objective algorithm for fuzzy flexible job shop with energy and transportation,” IEEE Transactions on Automation Science and Engi- neering, vol. 21, no. 3, pp. 4686–4702, 2024. JOURNAL OF LATEX CLASS FILES 9

    work page 2024

Show all 63 references
  1. [8]

    A decomposition-based evolutionary algorithm with clustering and hierarchical estimation for multi-objective fuzzy flexible jobshop scheduling,

    X. Zhang, S. Liu, Z. Zhao, and S. Yang, “A decomposition-based evolutionary algorithm with clustering and hierarchical estimation for multi-objective fuzzy flexible jobshop scheduling,” IEEE Transac- tions on Evolutionary Computation , early, access, 2024. doi: doi: 10.1109/TE...

    2024

  2. [9]

    Solving fuzzy job-shop schedul- ing problem using DE algorithm improved by a selection mechanism,

    D. Gao, G.-G. Wang, and W. Pedrycz, “Solving fuzzy job-shop schedul- ing problem using DE algorithm improved by a selection mechanism,” IEEE Transactions on Fuzzy Systems , vol. 28, no. 12, pp. 3265–3275, 2020

    2020

  3. [10]

    Improved artificial immune system algorithm for type-2 fuzzy flexible job shop scheduling problem,

    J.-q. Li, Z.-m. Liu, C. Li, and Z.-x. Zheng, “Improved artificial immune system algorithm for type-2 fuzzy flexible job shop scheduling problem,” IEEE Transactions on Fuzzy Systems , vol. 29, no. 11, pp. 3234–3248, 2021

    2021

  4. [12]

    Solving multiobjective fuzzy job-shop scheduling problem by a hybrid adaptive differential evolution algorithm,

    G.-G. Wang, D. Gao, and W. Pedrycz, “Solving multiobjective fuzzy job-shop scheduling problem by a hybrid adaptive differential evolution algorithm,” IEEE Transactions on Industrial Informatics, vol. 18, no. 12, pp. 8519–8528, 2022

    2022

  5. [13]

    A bi-population evolutionary algorithm with feedback for energy-efficient fuzzy flexible job shop scheduling,

    Z. Pan, D. Lei, and L. Wang, “A bi-population evolutionary algorithm with feedback for energy-efficient fuzzy flexible job shop scheduling,” IEEE Transactions on Systems, Man, and Cybernetics: Systems , vol. 52, no. 8, pp. 5295–5307, 2022

    2022

  6. [14]

    A learning-based memetic algorithm for energy-efficient flexible job-shop scheduling with type-2 fuzzy processing time,

    R. Li, W. Gong, C. Lu, and L. Wang, “A learning-based memetic algorithm for energy-efficient flexible job-shop scheduling with type-2 fuzzy processing time,” IEEE Transactions on Evolutionary Computa- tion, vol. 27, no. 3, pp. 610–620, 2023

    2023

  7. [15]

    A hyperheuristic with Q-learning for the multiobjective energy-efficient distributed blocking flow shop scheduling problem,

    F. Zhao, S. Di, and L. Wang, “A hyperheuristic with Q-learning for the multiobjective energy-efficient distributed blocking flow shop scheduling problem,” IEEE Transactions on Cybernetics , vol. 53, no. 5, pp. 3337– 3350, 2023

    2023

  8. [16]

    A reinforcement-learning-based 3-D estimation of distribution algorithm for fuzzy distributed hybrid flow- shop scheduling considering on-time-delivery,

    L. Deng, Y . Di, and L. Wang, “A reinforcement-learning-based 3-D estimation of distribution algorithm for fuzzy distributed hybrid flow- shop scheduling considering on-time-delivery,” IEEE Transactions on Cybernetics, vol. 54, no. 2, pp. 1024–1036, 2024

    2024

  9. [17]

    Co-evolution with deep reinforcement learning for energy-aware distributed heterogeneous flexible job shop scheduling,

    R. Li, W. Gong, L. Wang, C. Lu, and C. Dong, “Co-evolution with deep reinforcement learning for energy-aware distributed heterogeneous flexible job shop scheduling,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 54, no. 1, pp. 201–211, 2024

    2024

  10. [18]

    MQL-MM: A meta-Q-learning- based multi-objective metaheuristic for energy-efficient distributed fuzzy hybrid blocking flow-shop scheduling problem,

    Z. Shao, W. Shao, J. Chen, and D. Pi, “MQL-MM: A meta-Q-learning- based multi-objective metaheuristic for energy-efficient distributed fuzzy hybrid blocking flow-shop scheduling problem,” IEEE Transactions on Evolutionary Computation , early, access, 2024. 10.1109/TEVC.202 4.3399314

    2024 doi

  11. [19]

    A reinforcement learning driven cooperative meta-heuristic algorithm for energy-efficient distributed no- wait flow-shop scheduling with sequence-dependent setup time,

    F. Zhao, T. Jiang, and L. Wang, “A reinforcement learning driven cooperative meta-heuristic algorithm for energy-efficient distributed no- wait flow-shop scheduling with sequence-dependent setup time,” IEEE Transactions on Industrial Informatics , vol. 19, no. 7, pp. 8427–8440, 2023

    2023

  12. [20]

    A learning-based multipopulation evolutionary optimization for flexible job shop schedul- ing problem with finite transportation resources,

    Z. Pan, L. Wang, J. Zheng, J.-F. Chen, and X. Wang, “A learning-based multipopulation evolutionary optimization for flexible job shop schedul- ing problem with finite transportation resources,” IEEE Transactions on Evolutionary Computation, vol. 27, no. 6, pp. 1590–1603, 2023

    2023

  13. [21]

    An improved artificial bee colony algorithm with q-learning for solving permutation flow-shop scheduling problems,

    H. Li, K. Gao, P.-Y . Duan, J.-Q. Li, and L. Zhang, “An improved artificial bee colony algorithm with q-learning for solving permutation flow-shop scheduling problems,” IEEE Transactions on Systems, Man, and Cybernetics: Systems , vol. 53, no. 5, pp. 2684–2693, 2023

    2023

  14. [22]

    Learning-based grey wolf optimizer for stochastic flexible job shop scheduling,

    C. Lin, Z. Cao, and M. Zhou, “Learning-based grey wolf optimizer for stochastic flexible job shop scheduling,” IEEE Transactions on Automation Science and Engineering , vol. 19, no. 4, pp. 3659–3671, 2022

    2022

  15. [23]

    Learning to dispatch for job shop scheduling via deep reinforcement learning,

    C. Zhang, W. Song, Z. Cao, J. Zhang, P. S. Tan, and X. Chi, “Learning to dispatch for job shop scheduling via deep reinforcement learning,” in Advances in neural information processing systems , vol. 33, 2020, pp. 1621–1632

    2020

  16. [24]

    Fast approximations for job shop scheduling: A lagrangian dual deep learning method,

    J. Kotary, F. Fioretto, and P. Van Hentenryck, “Fast approximations for job shop scheduling: A lagrangian dual deep learning method,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 36, no. 7, 2022, pp. 7239–7246

    2022

  17. [25]

    Self-labeling the job shop scheduling problem,

    A. Corsini, A. Porrello, S. Calderara, and M. Dell’Amico, “Self-labeling the job shop scheduling problem,” arXiv preprint arXiv:2401.11849 , 2024

    2024 arXiv

  18. [26]

    Flexible job shop scheduling via dual attention network-based reinforcement learning,

    R. Wang, G. Wang, J. Sun, F. Deng, and J. Chen, “Flexible job shop scheduling via dual attention network-based reinforcement learning,” IEEE Transactions on Neural Networks and Learning Systems , vol. 35, no. 3, pp. 3091–3102, 2024

    2024

  19. [27]

    Deep reinforcement learning based optimization algorithm for permutation flow-shop scheduling,

    Z. Pan, L. Wang, J. Wang, and J. Lu, “Deep reinforcement learning based optimization algorithm for permutation flow-shop scheduling,” IEEE Transactions on Emerging Topics in Computational Intelligence , vol. 7, no. 4, pp. 983–994, 2023

    2023

  20. [28]

    An end-to-end reinforcement learning approach for job-shop scheduling problems based on constraint programming,

    P. Tassel, M. Gebser, and K. Schekotihin, “An end-to-end reinforcement learning approach for job-shop scheduling problems based on constraint programming,” in Proceedings of the International Conference on Au- tomated Planning and Scheduling , vol. 33, 2023, pp. 614–622

    2023

  21. [29]

    Flexible job-shop scheduling via graph neural network and deep reinforcement learning,

    W. Song, X. Chen, Q. Li, and Z. Cao, “Flexible job-shop scheduling via graph neural network and deep reinforcement learning,” IEEE Transactions on Industrial Informatics , vol. 19, no. 2, pp. 1600–1610, 2023

    2023

  22. [30]

    Dynamic parallel machine scheduling with deep q-network,

    C.-L. Liu, C.-J. Tseng, T.-H. Huang, and J.-W. Wang, “Dynamic parallel machine scheduling with deep q-network,” IEEE Transactions on Systems, Man, and Cybernetics: Systems , vol. 53, no. 11, pp. 6792– 6804, 2023

    2023

  23. [31]

    A knowledge-guided end- to-end optimization framework based on reinforcement learning for flow shop scheduling,

    Z. Pan, L. Wang, C. Dong, and J.-F. Chen, “A knowledge-guided end- to-end optimization framework based on reinforcement learning for flow shop scheduling,” IEEE Transactions on Industrial Informatics , vol. 20, no. 2, pp. 1853–1861, 2024

    2024

  24. [32]

    A deep reinforcement learning framework based on an attention mechanism and disjunctive graph embedding for the job-shop scheduling problem,

    R. Chen, W. Li, and H. Yang, “A deep reinforcement learning framework based on an attention mechanism and disjunctive graph embedding for the job-shop scheduling problem,” IEEE Transactions on Industrial Informatics, vol. 19, no. 2, pp. 1322–1331, 2023

    2023

  25. [33]

    A reinforcement learning approach for flexible job shop scheduling problem with crane transportation and setup times,

    Y . Du, J. Li, C. Li, and P. Duan, “A reinforcement learning approach for flexible job shop scheduling problem with crane transportation and setup times,” IEEE Transactions on Neural Networks and Learning Systems , vol. 35, no. 4, pp. 5695–5709, 2024

    2024

  26. [34]

    Large- scale dynamic scheduling for flexible job-shop with random arrivals of new jobs by hierarchical reinforcement learning,

    K. Lei, P. Guo, Y . Wang, J. Zhang, X. Meng, and L. Qian, “Large- scale dynamic scheduling for flexible job-shop with random arrivals of new jobs by hierarchical reinforcement learning,” IEEE Transactions on Industrial Informatics, vol. 20, no. 1, pp. 1007–1018, 2024

    2024

  27. [35]

    Double dqn- based coevolution for green distributed heterogeneous hybrid flowshop scheduling with multiple priorities of jobs,

    R. Li, W. Gong, L. Wang, C. Lu, Z. Pan, and X. Zhuang, “Double dqn- based coevolution for green distributed heterogeneous hybrid flowshop scheduling with multiple priorities of jobs,” IEEE Transactions on Automation Science and Engineering , vol. 21, no. 4, pp. 6550–6562, 2024

    2024

  28. [36]

    Integrating ma- chine learning and mathematical optimization for job shop scheduling,

    A. Liu, P. B. Luh, K. Sun, M. A. Bragin, and B. Yan, “Integrating ma- chine learning and mathematical optimization for job shop scheduling,” IEEE Transactions on Automation Science and Engineering , vol. 21, no. 3, pp. 4829–4850, 2024

    2024

  29. [37]

    Dynamic job-shop scheduling problems using graph neural network and deep reinforcement learning,

    C.-L. Liu and T.-H. Huang, “Dynamic job-shop scheduling problems using graph neural network and deep reinforcement learning,” IEEE Transactions on Systems, Man, and Cybernetics: Systems , vol. 53, no. 11, pp. 6836–6848, 2023

    2023

  30. [38]

    Real-time scheduling for dynamic partial- no-wait multiobjective flexible job shop by deep reinforcement learning,

    S. Luo, L. Zhang, and Y . Fan, “Real-time scheduling for dynamic partial- no-wait multiobjective flexible job shop by deep reinforcement learning,” IEEE Transactions on Automation Science and Engineering , vol. 19, no. 4, pp. 3020–3038, 2022

    2022

  31. [39]

    Optimal foraging algorithm that incorporates fuzzy relative entropy for solving many-objective permutation flow shop scheduling problems,

    G.-Y . Zhu, C. Ding, and W.-B. Zhang, “Optimal foraging algorithm that incorporates fuzzy relative entropy for solving many-objective permutation flow shop scheduling problems,” IEEE Transactions on Fuzzy Systems, vol. 28, no. 11, pp. 2738–2746, 2020

    2020

  32. [40]

    Fuzzy job-shop scheduling based on ranking level (/spl lambda/, 1) interval-valued fuzzy numbers,

    F.-T. Lin, “Fuzzy job-shop scheduling based on ranking level (/spl lambda/, 1) interval-valued fuzzy numbers,”IEEE Transactions on Fuzzy Systems, vol. 10, no. 4, pp. 510–522, 2002

    2002

  33. [41]

    Jobshop scheduling with imprecise durations: a fuzzy approach,

    P. Fortemps, “Jobshop scheduling with imprecise durations: a fuzzy approach,” IEEE Transactions on Fuzzy Systems , vol. 5, no. 4, pp. 557– 569, 1997

    1997

  34. [42]

    H. T. Nguyen, C. Walker, and E. A. Walker, A first course in fuzzy logic. Chapman and Hall/CRC, 2018

    2018

  35. [43]

    Fuzzy job shop scheduling problem with availability con- straints,

    D. Lei, “Fuzzy job shop scheduling problem with availability con- straints,” Computers & Industrial Engineering , vol. 58, no. 4, pp. 610– 617, 2010

    2010

  36. [44]

    An efficient genetic algorithm for job-shop scheduling problems with fuzzy processing time and fuzzy duedate,

    M. Sakawa and T. Mori, “An efficient genetic algorithm for job-shop scheduling problems with fuzzy processing time and fuzzy duedate,” Computers & industrial engineering , vol. 36, no. 2, pp. 325–341, 1999

    1999

  37. [45]

    The expected value of a fuzzy number,

    S. Heilpern, “The expected value of a fuzzy number,” Fuzzy sets and Systems, vol. 47, no. 1, pp. 81–86, 1992

    1992

  38. [46]

    Co- evolutionary makespan optimisation through different ranking methods for the fuzzy flexible job shop,

    J. J. Palacios, I. Gonz ´alez-Rodr´ıguez, C. R. Vela, and J. Puente, “Co- evolutionary makespan optimisation through different ranking methods for the fuzzy flexible job shop,” Fuzzy Sets and Systems , vol. 278, pp. 81–97, 2015

    2015

  39. [47]

    E. B. Tirkolaee, A. Goli, and G.-W. Weber, “Fuzzy mathematical programming and self-adaptive artificial fish swarm algorithm for just- in-time energy-aware flow shop scheduling problem with outsourcing JOURNAL OF LATEX CLASS FILES 10 option,” IEEE transactions on fuzzy systems...

    2020

  40. [48]

    Job shop scheduling by simulated annealing,

    P. J. Van Laarhoven, E. H. Aarts, and J. K. Lenstra, “Job shop scheduling by simulated annealing,” Operations research, vol. 40, no. 1, pp. 113– 125, 1992

    1992

  41. [49]

    How attentive are graph attention networks?

    S. Brody, U. Alon, and E. Yahav, “How attentive are graph attention networks?” arXiv preprint arXiv:2105.14491 , 2021

    2021 arXiv

  42. [50]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” in Advances in Neural Information Processing Systems , 2017

    2017

  43. [51]

    Understanding the difficulty of training deep feedforward neural networks,

    X. Glorot and Y . Bengio, “Understanding the difficulty of training deep feedforward neural networks,” in Proceedings of the thirteenth international conference on artificial intelligence and statistics . JMLR Workshop and Conference Proceedings, 2010, pp. 249–256

    2010

  44. [52]

    Maximum likelihood from incomplete data via the EM algorithm,

    A. P. Dempster, N. M. Laird, and D. B. Rubin, “Maximum likelihood from incomplete data via the EM algorithm,” Journal of the royal statistical society: series B (methodological) , vol. 39, no. 1, pp. 1–22, 1977

    1977

  45. [53]

    Many-objective evolutionary algorithm with reference point-based fuzzy correlation entropy for energy-efficient job shop scheduling with limited workers,

    W. Li, L. He, and Y . Cao, “Many-objective evolutionary algorithm with reference point-based fuzzy correlation entropy for energy-efficient job shop scheduling with limited workers,” IEEE Transactions on Cybernetics, vol. 52, no. 10, pp. 10 721–10 734, 2022

    2022

  46. [54]

    Fuzzy job-shop scheduling problems: A review,

    S. Abdullah and M. Abdolrazzagh-Nezhad, “Fuzzy job-shop scheduling problems: A review,”Information Sciences, vol. 278, pp. 380–407, 2014

    2014

  47. [55]

    Adam: A method for stochastic optimization,

    D. P. Kingma, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  48. [56]

    Mitchell, An introduction to genetic algorithms

    M. Mitchell, An introduction to genetic algorithms . MIT press, 1998

    1998

  49. [57]

    Particle swarm optimization,

    J. Kennedy and R. Eberhart, “Particle swarm optimization,” in Proceed- ings of ICNN’95-international conference on neural networks , vol. 4. ieee, 1995, pp. 1942–1948

    1995

  50. [58]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” in Advances in neural information processing systems, vol. 33, 2020, pp. 6840–6851

    2020

  51. [59]

    Generative adversarial nets,

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio, “Generative adversarial nets,” in Advances in neural information processing systems , vol. 27, 2014

    2014

  52. [60]

    Nice: Non-linear independent components estimation,

    L. Dinh, D. Krueger, and Y . Bengio, “Nice: Non-linear independent components estimation,” arXiv preprint arXiv:1410.8516 , 2014

    2014 arXiv

  53. [61]

    Permutation flow shop schedul- ing with multiple lines and demand plans using reinforcement learning,

    J. Brammer, B. Lutz, and D. Neumann, “Permutation flow shop schedul- ing with multiple lines and demand plans using reinforcement learning,” European Journal of Operational Research , vol. 299, no. 1, pp. 75–86, 2022

    2022

  54. [62]

    A tree-based multiobjective evo- lutionary algorithm for energy-efficient hybrid flow-shop scheduling,

    Y .-J. Wang, G.-G. Wang, and L. Wang, “A tree-based multiobjective evo- lutionary algorithm for energy-efficient hybrid flow-shop scheduling,” IEEE Transactions on Evolutionary Computation , early access, 2024. doi: 10.1109/TEVC.2024.3411661

    2024

  55. [63]

    A biobjective optimization for integrated parallel machine scheduling and location problem: Mathe- matical model and iterative two-stage heuristic,

    P. Wu, Y . Wang, J. Cheng, and Y . Li, “A biobjective optimization for integrated parallel machine scheduling and location problem: Mathe- matical model and iterative two-stage heuristic,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 53, no. 11, pp. 6897–6...

    2023

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.