Pith. sign in

REVIEW 2 major objections 4 minor 1 cited by

Multi-task Representation Learning for Mixed Integer Linear Programming

T0 review · 2 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A single shared MILP embedding trained on two solver-guidance tasks can be fine-tuned to a third task and to larger instances, beating models trained on that task alone.

desk verdict Useful multi-task pretraining for MILP with a real confound: gains may come from extra data, not multi-task learning per se. read the letter →

arxiv 2412.14409 v2 pith:74RPN77S submitted 2024-12-18 cs.AI cs.LGmath.OC

Junyang Cai , Taoan Huang , Bistra Dilkina This is my paper · ORCID
classification cs.AIcs.LGmath.OC
keywords multi-tasklearningmixedintegerlinearprogramminggraphattentionnetworkscontrastivebackdoorspredict-and-searchsolverconfigurationgeneralization
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

The paper sets out to show that MILP solvers do not need a separate trained model for every decision task. It proposes a multi-task training scheme in which one shared graph-attention encoder learns a MILP embedding by alternating between two tasks, while randomly initialised task-specific heads stay frozen; afterwards the heads are fine-tuned and the encoder is not touched. The claim is that this shared embedding matches task-specific models on training-distribution instances and generalises better to larger instances and to a third, unseen task, across three benchmarks and two solvers. If true, it means a single pretrained MILP representation can be cheaply adapted to new solver-guidance tasks, reducing the data collection and training cost of ML-guided MILP solving.

What carries the argument

The core mechanism is a two-step training schedule. In step one, the shared graph-attention encoder processes each MILP as a bipartite graph of variable and constraint nodes and is trained with the InfoNCE contrastive loss on alternating batches from two tasks, while three randomly initialised task heads are kept frozen; in step two, the task heads are fine-tuned with the encoder fixed, so the shared embedding is not distorted by any one task. This separation, together with the permutation-invariant bipartite encoding, is what lets the same embedding feed branching-priority prediction, solution-neighborhood prediction, and solver-configuration generation.

What would settle it

Train a single-task model on the combined data of the two pretraining tasks, using the same number of samples and the same encoder architecture, and then fine-tune it on the third task; if that model matches the multi-task model's performance on large instances and cross-task transfer, the claimed benefit of shared task structure is not needed.

Watch

Extended reading notes

Core claim

The paper's central claim is that multi-task representation learning provides a reusable MILP embedding: a graph-attention encoder trained jointly on two tasks yields a representation that, after fine-tuning only a small task-specific head, performs at least as well as a fully trained single-task model on the same task and clearly better when the test instances are larger or the task is new. This is demonstrated on Combinatorial Auction, Maximum Independent Set, and Minimum Vertex Cover, under both Gurobi and SCIP, with the shared model matching or improving solve time, primal integral, and solution quality compared with single-task models.

Load-bearing premise

The decisive assumption is that the gains come from sharing representations between tasks, but the experiments do not compare against a single-task model trained on the same total amount of data, so the observed advantage might simply reflect more training data rather than multi-task learning.

Editorial extensions

If this is right

  • A model trained on two MILP-guidance tasks can be adapted to a third by training only a small task-specific head, with fine-tuning converging in under an hour in the experiments.
  • On larger instances, the multi-task model improves average solve time for Backdoors by roughly 15% over Gurobi and reduces PaS primal integral by up to about 92% over Gurobi, while the single-task models gain less or degrade.
  • The same shared embedding helps a new Configuration task generate per-instance SCIP settings that beat both SMAC and single-task configuration models, especially on large instances.
  • Cross-task checks, where each possible pair is trained and the third task is fine-tuned, all show improvements over single-task models, suggesting the result is not tied to one pair of tasks.
  • Because the framework is stated to work with any MILP graph representation, size-invariant network, and loss function, it can be slotted onto other solver-guidance tasks without redesign.

Reading between the lines

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

  • The paper's attribution of the gains to multi-task learning is not fully isolated: no single-task pretraining baseline on the same data volume appears in the experiments, so a reader should treat the mechanism as plausible but not proven.
  • The fast fine-tuning suggests that if larger pretrained MILP encoders are built, adapting them to new solver tasks could cost far less than training from scratch; this is the direction the paper gestures toward with its mention of foundation models for MILP.
  • The Configuration head shows a lightweight way to produce solver parameters directly per instance, which could be tested on other solvers and on a broader parameter space than the 15 SCIP parameters used here.
  • The three tasks all rely on one-time inference from static instance features; extending the same shared embedding to dynamic features such as node selection during branch-and-bound is a natural but untested next step.
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

2 major / 4 minor

Summary. The paper introduces a multi-task representation learning framework for MILP solving. A shared GAT encoder is trained on two tasks (Backdoor, Predict-and-Search, or Solver Configuration) using a two-step procedure: first training the shared architecture with fixed, randomly initialized task-specific heads, then fine-tuning the task-specific heads for a target task. Experiments on Combinatorial Auction, Maximal Independent Set, and Minimum Vertex Cover benchmarks compare multi-task models against single-task models and default solvers. The reported results show that multi-task models are competitive with single-task models on in-distribution small instances and often superior on larger instances and on a new task, leading the authors to claim that multi-task representation learning improves generalization.

Significance. If the central claim is established, the framework could reduce the training cost for new MILP-guided tasks by reusing a shared embedding, and it could improve generalization to larger instances. The paper provides a public code repository and evaluates on genuinely held-out larger instances and cross-task fine-tuning, which are strengths. However, the headline attribution of the observed gains to multi-task learning is not yet supported because the multi-task and single-task comparisons differ in confounding factors, as detailed in the major comments.

major comments (2)
  1. [§5.1 (Baselines and Approaches) and §5.2 (Same-Task Performance, New-Task Performance)] The comparison between Multi-task and Single-task confounds multi-task learning with pre-training on additional data. Single-task is trained on 200 instances of the target task only, while Multi-task first trains the shared encoder on 400 instances (200 from each of two tasks) and then fine-tunes on the target task's 200 instances. Thus, the reported gains on large instances and on new tasks could be caused by seeing twice as many training instances, or by any form of pre-training on a related task, rather than by the multi-task objective itself. The interpretation in §5.2 ('By learning a generalized embedding that avoids overfitting to a single distribution…') is asserted without isolating this mechanism. To support the paper's central claim, the authors should add a single-task pre-training control: pre-train the same architecture on one task (e.g., Backdoor only) using the same two-step procedure and the same total data volume (200 + 200 instances), then fine-tune on the target task. If this control performs as well as Multi-task, the multi-task attribution would not be supported; if not, it would substantiate the claim.
  2. [Table 2, MIS-L PaS row] The Single-task PaS model achieves a Primal Integral of 50.83, which is substantially worse than the Gurobi baseline (30.73), while Multi-task-BAPAS achieves 2.23. Such a collapse suggests that the Single-task baseline may be under-trained or unstable for this distribution, which would inflate the apparent multi-task advantage. The paper does not report training curves or multiple random seeds for the Single-task baselines. To ensure a fair comparison, the authors should show that Single-task models are trained to convergence and report variance across seeds, or discuss any hyperparameter differences that might explain this failure.
minor comments (4)
  1. [Table 3] The column header says 'Gurobi' for all task rows, but the Configuration rows actually compare SCIP, not Gurobi. This mislabeling should be corrected.
  2. [§4 (Multi-task Representation Learning)] The text says 'We use three randomly initialized task-specific layers to enhance the robustness of each task,' which is ambiguous: it is unclear whether three heads are used per task or three heads total across the two tasks. Please clarify.
  3. [§4.2 (Network architecture and Contrastive loss)] The contrastive loss formula appears to have a missing closing parenthesis in the denominator: the expression `exp(a′⊤pθ(P)/τ )` should likely be `exp(a′⊤pθ(P)/τ)`. Please check the typesetting.
  4. [§1 (Introduction)] The claim 'first multi-task learning framework for ML-guided MILP solving' may be too strong given that multi-task learning is common in other domains; the related work section does acknowledge this, but the introduction should qualify the claim to avoid overstatement.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the held-out evaluations and cross-task fine-tuning are not constructed from the target claims.

full rationale

The paper's central claim is that a shared MILP embedding trained on two tasks, then fine-tuned on a third task or transferred to larger instances, generalizes better than a task-specific model trained from scratch. This claim is supported by genuinely held-out evidence: large instances (L) are not used in training or fine-tuning, and cross-task evaluations fine-tune on the target task using the same 200 instances as the Single-task baseline. The model's predictions are evaluated against solver performance metrics (solve time, primal gap, primal integral) that are external to the training objective and not defined in terms of the claimed conclusion. The contrastive loss and data collection procedures are taken from prior work, including the authors' own earlier papers, but those prior components supply the architecture and training signal; they do not encode the generalization result. The most serious concern is a confound, not circularity: the Multi-task setting differs from Single-task both in task diversity and in the quantity of pretraining data, so the observed gains could be attributable to pretraining/data volume rather than to multi-task representation learning specifically. A single-task pretraining control would disambiguate this, and the paper's Section 5.2 interpretation ('By learning a generalized embedding...') is asserted rather than isolated by such a control. However, this is an experimental attribution gap, not a case where a prediction reduces by construction to its inputs, a fitted parameter is renamed as a prediction, or a load-bearing conclusion rests on a self-citation. No circular step can be exhibited from the paper's equations or definitions; therefore the circularity score is 0.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The framework relies on standard ML assumptions: a bipartite graph with hand-crafted features is sufficient, contrastive InfoNCE with tau=0.07 is effective, the external data generators (MCTS, SMAC3, PaS) provide valid labels, and the two-step random-head training yields a transferable embedding. None of these is new to this paper; they are inherited from the cited prior work.

assumptions (5)
  • domain assumption The bipartite graph representation with 15 variable features, 4 constraint features, and 1 edge feature captures enough structure for all three tasks.
    Introduced in Section 4.1; the feature set is taken from prior work [21] without a study of whether it is sufficient for multi-task sharing.
  • domain assumption Contrastive InfoNCE loss with temperature tau=0.07 provides effective supervision for backdoor, PaS, and configuration tasks.
    Section 4.2; adopted from [28] and applied unchanged to all tasks.
  • domain assumption Positive and negative sample generation (MCTS backdoors, near-optimal PaS solutions plus 10-percent-neighborhood negative solutions, SMAC3 configurations) yields reliable training labels.
    Section 4.1; the paper relies on external collection pipelines without validating label quality for each task.
  • ad hoc to paper The two-step training strategy with randomly initialized fixed task-specific layers produces a shared embedding that transfers across tasks.
    Section 4; this is the paper's own design choice, not independently justified by an ablation or theory.
  • domain assumption Large test instances are drawn from the same distribution as small training instances except for size.
    Section 5.1, Table 1; instance generators keep density fixed while increasing nodes or bids.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multi-task Representation Learning for Mixed Integer Linear Programming." pith.science (2026). https://pith.science/paper/74RPN77S

@misc{pith2026241214409,
  author       = {Pith},
  title        = {Pith review of: Multi-task Representation Learning for Mixed Integer Linear Programming},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/74RPN77S}},
  note         = {Machine review of arXiv:2412.14409}
}
read the original abstract

Mixed Integer Linear Programs (MILPs) are highly flexible and powerful tools for modeling and solving complex real-world combinatorial optimization problems. Recently, machine learning (ML)-guided approaches have demonstrated significant potential in improving MILP-solving efficiency. However, these methods typically rely on separate offline data collection and training processes, which limits their scalability and adaptability. This paper introduces the first multi-task learning framework for ML-guided MILP solving. The proposed framework provides MILP embeddings helpful in guiding MILP solving across solvers (e.g., Gurobi and SCIP) and across tasks (e.g., Branching and Solver configuration). Through extensive experiments on three widely used MILP benchmarks, we demonstrate that our multi-task learning model performs similarly to specialized models within the same distribution. Moreover, it significantly outperforms them in generalization across problem sizes and tasks.

Figures

Figures reproduced from arXiv: 2412.14409 by the authors.

Figure 1
Figure 1. This figure illustrates the multi-task learning framework for Mixed Integer [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Same-Task Performance: The Primal Gap (the lower, the better) as a runtime function averaged over 100 test instances on PaS and each benchmark. We compare the performance of Gurobi (green line), Single-task on PaS (blue line), and Multi-task-BAPAS (orange line). – Can the model, with its shared MILP embeddings, effectively fine-tune task￾specific layers to handle new tasks, and how does its performance compare to bo… view at source ↗
Figure 3
Figure 3. New-task Performance: The Primal Gap (the lower, the better) as a runtime function averaged over 100 test instances on Configuration and each benchmark. We compare the performance of SCIP (green line), SMAC (red line), Single-task on Configuration (blue line), and Multi-task-BAPAS (orange line). The results differ significantly when evaluating generalization performance on 100 L instances (directly tested without tr… view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Balans: Multi-Armed Bandits-based Adaptive Large Neighborhood Search for Mixed-Integer Programming Problem

    cs.AI 2024-12 conditional novelty 6.0 of 10

    Balans uses online multi-armed bandits to adaptively choose among large-neighborhood search operators on top of a MIP solver, reporting large primal gap improvements over default SCIP and Gurobi on hard instances with...

Reference graph

Works this paper leans on

57 extracted references · 37 canonical work pages · cited by 1 Pith paper

  1. [1]

    In: Operations Research Proceedings 2011: Selected Papers of the International Conference on Operations Research (OR 2011), August 30-September 2, 2011, Zurich, Switzerland

    Achterberg, T., Berthold, T., Hendel, G.: Rounding and propagation heuristics for mixed integer programming. In: Operations Research Proceedings 2011: Selected Papers of the International Conference on Operations Research (OR 2011), August 30-September 2, 2011, Zurich, Switzerland. pp. 71–76. Springer (2012)

    2012

  2. [2]

    Reviews of modern physics 74(1), 47 (2002)

    Albert, R., Barabási, A.L.: Statistical mechanics of complex networks. Reviews of modern physics 74(1), 47 (2002)

    work page 2002

  3. [3]

    INFORMS Journal on Computing29(1), 185–195 (2017)

    Alvarez, A.M., Louveaux, Q., Wehenkel, L.: A machine learning-based approxima- tion of strong branching. INFORMS Journal on Computing29(1), 185–195 (2017)

    work page 2017

  4. [4]

    Berthold, T.: Primal heuristics for mixed integer programs. Ph.D. thesis, Zuse Institute Berlin (ZIB) (2006)

    2006

  5. [5]

    arXiv preprint arXiv:2402.17702 (2024)

    Bolusani, S., Besançon, M., Bestuzheva, K., Chmiela, A., Dionísio, J., Donkiewicz, T., van Doornmalen, J., Eifler, L., Ghannam, M., Gleixner, A., et al.: The scip optimization suite 9.0. arXiv preprint arXiv:2402.17702 (2024)

    arXiv 2024

  6. [6]

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

    arXiv 2021

  7. [7]

    arXiv preprint arXiv:2401.10467 (2024)

    Cai, J., Huang, T., Dilkina, B.: Learning backdoors for mixed integer programs with contrastive learning. arXiv preprint arXiv:2401.10467 (2024)

    arXiv 2024

  8. [8]

    Balans: Multi-Armed Bandits-based Adaptive Large Neighborhood Search for Mixed-Integer Programming Problem

    Cai, J., Kadioglu, S., Dilkina, B.: Balans: Multi-armed bandits-based adaptive large neighborhood search for mixed-integer programming problem. arXiv preprint arXiv:2412.14382 (2024)

    work page Pith review arXiv 2024

Show all 57 references
  1. [9]

    In: International Conference on Integration of Constraint Programming, Artificial Intelligence, and Operations Research

    Cai, J., Nguyen, K.N., Shrestha, N., Good, A., Tu, R., Yu, X., Zhe, S., Serra, T.: Getting away with more network pruning: From sparsity to geometry and linear regions. In: International Conference on Integration of Constraint Programming, Artificial Intelligence, and Operatio...

    2023

  2. [10]

    arXiv preprint arXiv:2310.09478 (2023)

    Chen, J., Zhu, D., Shen, X., Li, X., Liu, Z., Zhang, P., Krishnamoorthi, R., Chan- dra, V., Xiong, Y., Elhoseiny, M.: Minigpt-v2: large language model as a unified interface for vision-language multi-task learning. arXiv preprint arXiv:2310.09478 (2023)

    2023 arXiv

  3. [11]

    Advances in Neural Information Processing Systems 34, 24235–24246 (2021)

    Chmiela, A., Khalil, E., Gleixner, A., Lodi, A., Pokutta, S.: Learning to sched- ule heuristics in branch and bound. Advances in Neural Information Processing Systems 34, 24235–24246 (2021)

    2021

  4. [12]

    In: CPAIOR

    Dilkina, B., Gomes, C.P.: Solving connected subgraph problems in wildlife conser- vation. In: CPAIOR. vol. 6140, pp. 102–116. Springer (2010)

    2010

  5. [13]

    Dilkina, B., Gomes, C.P., Malitsky, Y., Sabharwal, A., Sellmann, M.: Backdoors to combinatorial optimization: Feasibility and optimality. In: Integration of AI and OR Techniques in Constraint Programming for Combinatorial Optimization Problems: 6th International Conference, CP...

    2009

  6. [14]

    In: Proceedings of the aaai conference on artificial intelligence

    Ding, J.Y., Zhang, C., Shen, L., Li, S., Wang, B., Xu, Y., Song, L.: Accelerating primal solution findings for mixed integer programs based on solution prediction. In: Proceedings of the aaai conference on artificial intelligence. vol. 34, pp. 1452– 1459 (2020)

    2020

  7. [15]

    Annals of mathematics pp

    Dinur, I., Safra, S.: On the hardness of approximating minimum vertex cover. Annals of mathematics pp. 439–485 (2005)

    2005

  8. [16]

    arXiv e-prints pp

    Drakulic, D., Michel, S., Andreoli, J.M.: Goal: A generalist combinatorial opti- mization agent learner. arXiv e-prints pp. arXiv–2406 (2024) Multi-task Representation Learning for Mixed Integer Linear Programming 17

    2024

  9. [17]

    In: International Conference on Integration of Constraint Pro- gramming, Artificial Intelligence, and Operations Research

    Ferber, A., Song, J., Dilkina, B., Yue, Y.: Learning pseudo-backdoors for mixed integer programs. In: International Conference on Integration of Constraint Pro- gramming, Artificial Intelligence, and Operations Research. pp. 91–102. Springer (2022)

    2022

  10. [18]

    In: International Conference on In- teger Programming and Combinatorial Optimization

    Fischetti, M., Monaci, M.: Backdoor branching. In: International Conference on In- teger Programming and Combinatorial Optimization. pp. 183–191. Springer (2011)

    2011

  11. [19]

    Annals of Operations Research139, 131– 162 (2005)

    Floudas, C.A., Lin, X.: Mixed integer linear programming in process scheduling: Modeling, algorithms, and applications. Annals of Operations Research139, 131– 162 (2005)

    2005

  12. [20]

    In: NeurIPS 2021 competitions and demonstrations track

    Gasse, M., Bowly, S., Cappart, Q., Charfreitag, J., Charlin, L., Chételat, D., Chmiela, A., Dumouchelle, J., Gleixner, A., Kazachkov, A.M., et al.: The ma- chine learning for combinatorial optimization competition (ml4co): Results and insights. In: NeurIPS 2021 competitions an...

    2022

  13. [21]

    Advances in neural infor- mation processing systems32 (2019)

    Gasse, M., Chételat, D., Ferroni, N., Charlin, L., Lodi, A.: Exact combinatorial optimization with graph convolutional neural networks. Advances in neural infor- mation processing systems32 (2019)

    2019

  14. [22]

    Gurobi Optimization, LLC: Gurobi Optimizer Reference Manual (2024),https: //www.gurobi.com

    2024

  15. [23]

    arXiv preprint arXiv:2302.05636 (2023)

    Han, Q., Yang, L., Chen, Q., Zhou, X., Zhang, D., Wang, A., Sun, R., Luo, X.: A gnn-guided predict-and-search framework for mixed-integer linear programming. arXiv preprint arXiv:2302.05636 (2023)

    2023 arXiv

  16. [24]

    Advances in neural information processing systems27 (2014)

    He, H., Daume III, H., Eisner, J.M.: Learning to search in branch and bound algorithms. Advances in neural information processing systems27 (2014)

    2014

  17. [25]

    Mathematical Programming Computation14, 185–221 (2022)

    Hendel, G.: Adaptive large neighborhood search for mixed integer programming. Mathematical Programming Computation14, 185–221 (2022)

    2022

  18. [26]

    Annals of Operations Research339(1), 909–936 (2024)

    Hosny, A., Reda, S.: Automatic milp solver configuration by learning problem similarities. Annals of Operations Research339(1), 909–936 (2024)

    2024

  19. [27]

    In: Proceedings of the 3rd ACM SIGCAS Conference on Computing and Sustainable Societies

    Huang, T., Dilkina, B.: Enhancing seismic resilience of water pipe networks. In: Proceedings of the 3rd ACM SIGCAS Conference on Computing and Sustainable Societies. pp. 44–52 (2020)

    2020

  20. [28]

    In: International Conference on Machine Learning

    Huang, T., Ferber, A.M., Tian, Y., Dilkina, B., Steiner, B.: Searching large neigh- borhoods for integer linear programs with contrastive learning. In: International Conference on Machine Learning. pp. 13869–13890. PMLR (2023)

    2023

  21. [29]

    In: International Conference on Machine Learning

    Huang, T., Ferber, A.M., Zharmagambetov, A., Tian, Y., Dilkina, B.: Contrastive predict-and-search for mixed integer linear programs. In: International Conference on Machine Learning. PMLR (2024)

    2024

  22. [30]

    arXiv preprint arXiv:2406.06954 (2024)

    Huang, W., Huang, T., Ferber, A.M., Dilkina, B.: Distributional MIPLIB: a multi-domain library for advancing ml-guided milp methods. arXiv preprint arXiv:2406.06954 (2024)

    2024 arXiv

  23. [31]

    In: ECAI 2010, pp

    Kadioglu, S., Malitsky, Y., Sellmann, M., Tierney, K.: Isac–instance-specific algo- rithm configuration. In: ECAI 2010, pp. 751–756. IOS Press (2010)

    2010

  24. [32]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Khalil,E.,LeBodic,P.,Song,L.,Nemhauser,G.,Dilkina,B.:Learningtobranchin mixed integer programming. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 30 (2016)

    2016

  25. [33]

    In: Ijcai

    Khalil, E.B., Dilkina, B., Nemhauser, G.L., Ahmed, S., Shao, Y.: Learning to run heuristics in tree search. In: Ijcai. pp. 659–666 (2017)

    2017

  26. [34]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Khalil, E.B., Morris, C., Lodi, A.: Mip-gnn: A data-driven framework for guid- ing combinatorial solvers. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 36, pp. 10219–10227 (2022) 18 J.Cai, T.Huang, B.Dilkina

    2022

  27. [35]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Khalil, E.B., Vaezipoor, P., Dilkina, B.: Finding backdoors to integer programs: a monte carlo tree search framework. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 36, pp. 3786–3795 (2022)

    2022

  28. [36]

    arXiv preprint arXiv:1412.6980 (2014)

    Kingma, D.P., Ba, J.: Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014)

    2014 arXiv

  29. [37]

    Advances in Neural Information Processing Systems 35, 32000–32010 (2022)

    Labassi, A.G., Chételat, D., Lodi, A.: Learning to compare nodes in branch and bound with graph neural networks. Advances in Neural Information Processing Systems 35, 32000–32010 (2022)

    2022

  30. [38]

    Springer (2010)

    Land, A.H., Doig, A.G.: An automatic method for solving discrete programming problems. Springer (2010)

    2010

  31. [39]

    In: Proceedings of the 2nd ACM conference on Electronic commerce

    Leyton-Brown, K., Pearson, M., Shoham, Y.: Towards a universal test suite for combinatorial auction algorithms. In: Proceedings of the 2nd ACM conference on Electronic commerce. pp. 66–76 (2000)

    2000

  32. [40]

    arXiv preprint arXiv:2410.08288 (2024)

    Li, S., Kulkarni, J., Menache, I., Wu, C., Li, B.: Towards foundation models for mixed integer linear programming. arXiv preprint arXiv:2410.08288 (2024)

    2024 arXiv

  33. [41]

    arXiv preprint arXiv:2402.03647 (2024)

    Lin, J., Xu, M., Xiong, Z., Wang, H.: Cambranch: Contrastive learning with aug- mented milps for branching. arXiv preprint arXiv:2402.03647 (2024)

    2024 arXiv

  34. [42]

    Journal of Machine Learning Research 23(54), 1–9 (2022)

    Lindauer, M., Eggensperger, K., Feurer, M., Biedenkapp, A., Deng, D., Benjamins, C., Ruhkopf, T., Sass, R., Hutter, F.: Smac3: A versatile bayesian optimization package for hyperparameter optimization. Journal of Machine Learning Research 23(54), 1–9 (2022)

    2022

  35. [43]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recog- nition

    Liu, S., Johns, E., Davison, A.J.: End-to-end multi-task learning with attention. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recog- nition. pp. 1871–1880 (2019)

    2019

  36. [44]

    Top25, 207–236 (2017)

    Lodi, A., Zarpellon, G.: On learning and branching: a survey. Top25, 207–236 (2017)

    2017

  37. [45]

    arXiv preprint arXiv:2012.13349 (2020)

    Nair, V., Bartunov, S., Gimeno, F., Von Glehn, I., Lichocki, P., Lobov, I., O’Donoghue, B., Sonnerat, N., Tjandraatmadja, C., Wang, P., et al.: Solving mixed integer programs using neural networks. arXiv preprint arXiv:2012.13349 (2020)

    2020 arXiv

  38. [46]

    arXiv preprint arXiv:1807.03748 (2018)

    Oord, A.v.d., Li, Y., Vinyals, O.: Representation learning with contrastive predic- tive coding. arXiv preprint arXiv:1807.03748 (2018)

    2018 arXiv

  39. [47]

    In: International conference on machine learning

    Paulus, M.B., Zarpellon, G., Krause, A., Charlin, L., Maddison, C.: Learning to cut by looking ahead: Cutting plane selection via imitation learning. In: International conference on machine learning. pp. 17584–17600. PMLR (2022)

    2022

  40. [48]

    Artificial intelligence 1(3-4), 193–204 (1970)

    Pohl, I.: Heuristic search viewed as path finding in a graph. Artificial intelligence 1(3-4), 193–204 (1970)

    1970

  41. [49]

    Mathematical Program- ming pp

    Scavuzzo, L., Aardal, K., Lodi, A., Yorke-Smith, N.: Machine learning augmented branch and bound for mixed integer linear programming. Mathematical Program- ming pp. 1–44 (2024)

    2024

  42. [50]

    In: Annual Conference on Neural Infor- mation Processing Systems (NeurIPS) (2020)

    Song, J., Lanka, R., Yue, Y., Dilkina, B.: A general large neighborhood search framework for solving integer programs. In: Annual Conference on Neural Infor- mation Processing Systems (NeurIPS) (2020)

    2020

  43. [51]

    arXiv preprint arXiv:1804.00846 (2018)

    Song, J., Lanka, R., Zhao, A., Bhatnagar, A., Yue, Y., Ono, M.: Learning to search via retrospective imitation. arXiv preprint arXiv:1804.00846 (2018)

    2018 arXiv

  44. [52]

    In: International conference on machine learning

    Tang,Y.,Agrawal,S.,Faenza,Y.:Reinforcementlearningforintegerprogramming: Learning to cut. In: International conference on machine learning. pp. 9367–9376. PMLR (2020)

    2020

  45. [53]

    SIAM Jour- nal on Computing6, 537–546 (1977) Multi-task Representation Learning for Mixed Integer Linear Programming 19

    Tarjan, R.E., Trojanowski, A.E.: Finding a maximum independent set. SIAM Jour- nal on Computing6, 537–546 (1977) Multi-task Representation Learning for Mixed Integer Linear Programming 19

    1977

  46. [54]

    In: International Conference on the Integration of Constraint Pro- gramming, Artificial Intelligence, and Operations Research

    Tong, J., Cai, J., Serra, T.: Optimization over trained neural networks: Taking a relaxing walk. In: International Conference on the Integration of Constraint Pro- gramming, Artificial Intelligence, and Operations Research. pp. 221–233. Springer (2024)

    2024

  47. [55]

    arXiv preprint arXiv:2202.04910 (2022)

    Valentin, R., Ferrari, C., Scheurer, J., Amrollahi, A., Wendler, C., Paulus, M.B.: Instance-wise algorithm configuration with graph neural networks. arXiv preprint arXiv:2202.04910 (2022)

    2022 arXiv

  48. [56]

    In: IJCAI

    Williams, R., Gomes, C.P., Selman, B.: Backdoors to typical case complexity. In: IJCAI. vol. 3, pp. 1173–1178 (2003)

    2003

  49. [57]

    Xu, L., Hutter, F., Hoos, H.H., Leyton-Brown, K.: Hydra-mip: Automated al- gorithm configuration and selection for mixed integer programming. In: RCRA workshop on experimental evaluation of algorithms for solving problems with com- binatorial explosion at the international joi...

    2011

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.