Pith. sign in

REVIEW 3 major objections 4 minor 49 references

Optimization Learning

T0 review · 3 major / 4 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 The paper claims that optimization proxies—differentiable programs that combine a neural network with repair or completion layers—can learn the input-output map of a parametric optimization problem and return feasible, near-optimal…

desk verdict A clear, well-written synthesis of the author's own optimization-learning line, but the 'trustworthy by design' claim overreaches: the flagship SCOPF application gets feasibility from training, not architecture. read the letter →

arxiv 2501.03443 v1 pith:URWLF3GM submitted 2025-01-07 math.OC cs.AI

Pascal Van Hentenryck This is my paper · ORCID
classification math.OCcs.AI MSC 68T0790C31
keywords optimizationlearningproxiesparametricself-supervisedprimal-dualeconomicdispatchsecurity-constrainedoptimalpowerflowdifferentiableprogramming
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 is trying to establish that the repeated solving of a parametric optimization problem can be replaced by a trained differentiable program, called an optimization proxy, that outputs feasible points, while a companion dual proxy outputs a valid lower bound. This matters because many engineering optimization problems are solved over and over on nearly identical inputs, and exact solvers are too slow to run in real time or inside an interactive decision loop. If the proxy behaves as claimed, the same problem class can be answered in milliseconds with a certified bound, and end-to-end self-supervised training removes the need for precomputed optimal labels. The paper demonstrates the concept on power-system applications: economic dispatch for grids up to 30,000 buses, DC optimal power flow, and security-constrained optimal power flow under N-1 contingencies.

What carries the argument

The machinery is the repair/completion layer: a hard-coded, differentiable map that turns whatever the neural network predicts into a point that satisfies the problem's constraints. The generic template is the closest-point projection, $\min_y \|y - p_y\|$ subject to $h_x(y)=0$ and $g_x(y)\ge 0$, but the paper's practical examples are hand-derived closed forms: proportional rescaling for power balance and reserves in economic dispatch, positive/negative-part completion for linear-programming duals, and a binary-search layer for contingency dispatches in security-constrained OPF. Because each layer has subgradients almost everywhere, the original objective can be backpropagated through the whole model, making the composed system a differentiable program that is also fast at inference.

What would settle it

On a random batch of held-out instances from the same load distribution, evaluate the trained primal and dual proxies on the same grid; the central claim fails if any primal output violates a hard constraint, or if the primal objective falls below the dual objective, since that would mean the dual proxy is not a valid lower bound.

Watch

Extended reading notes

Core claim

The central claim is that parametric optimization can be outsourced to a proxy that is trustworthy by design: a trained composition of a neural network and a repair or completion layer is constructed to land in the feasible set, and a dual proxy supplies a certified lower bound. Unlike a regression model, the proxy is not merely close to a solution; its output satisfies the constraint system, and because the training loss is the original objective function, the model can be trained self-supervised, without solved labels. On economic dispatch with perturbed loads, the end-to-end primal proxy keeps optimality gaps mostly below 1% on grids up to 30,000 buses while evaluating a batch of 256 instances in roughly 10 ms. On DC optimal power flow, the dual proxy produces geometric-mean dual gaps around 0.25-0.5%. On security-constrained optimal power flow with N-1 generator and line contingencies, the primal-dual proxy returns near-optimal dispatches in about 10 ms per instance, four orders of magnitude faster than the commercial solver baseline on systems with up to 6,515 buses.

Load-bearing premise

Every problem class of interest must admit a repair or completion layer that is both fast and differentiable; the paper gives working examples for dispatch, linear programs, and conic cases, but no general recipe that this layer always exists.

Editorial extensions

If this is right

  • Real-time risk assessment becomes practical: evaluating a 24-hour Monte-Carlo scenario of five-minute dispatches drops from roughly 15 minutes of solver time to about 5 seconds of proxy evaluation.
  • Self-supervised training removes the label bottleneck: for the largest tested security-constrained OPF case the reported training time of under three hours on a GPU is less than the time a solver needs to produce optimal labels for a single hard instance.
  • A dual proxy gives a cheap quality certificate: on DC-OPF the geometric-mean dual gap is around 0.25-0.5%, so a per-instance lower bound can be produced in milliseconds, not minutes.
  • The architecture pattern, raw prediction plus differentiable repair/completion, transfers across economic dispatch, linear programs, conic relaxations, and security-constrained OPF, which is why the paper presents it as a general methodology rather than a single model.

Reading between the lines

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

  • A natural safety protocol the paper does not describe is to run the primal and dual proxies together and send any instance whose primal-dual gap exceeds a threshold back to an exact solver; most instances would still be served in milliseconds, and the exceptions would keep a formal guarantee.
  • The by-design feasibility guarantee is only as strong as the repair layer's coverage under distribution shift; a concrete stress test is to evaluate trained proxies on load patterns outside the training distribution and measure how the primal-dual gap degrades.
  • If automated synthesis of differentiable repair layers for general constraint classes were built, this approach would become a compiler-like technology; the paper itself lists the derivation of effective repair layers for a wide range of applications as an open issue.
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

3 major / 4 minor

Summary. This paper introduces 'optimization learning' as a framework for learning parametric optimization mappings with differentiable programs that combine neural networks with repair or completion layers. Three proxy architectures are presented: primal optimization proxies, dual optimization proxies, and primal-dual learning. The framework is illustrated on economic dispatch with reserve requirements, DC optimal power flow, and security-constrained optimal power flow, using PGLib benchmarks with up to 30,000 buses and reporting millisecond inference times and speedups of several orders of magnitude over commercial solvers. The central claim is that these proxies are 'trustworthy by design': they compute feasible solutions, provide quality guarantees, and scale to large instances.

Significance. If the claims were fully supported, the framework would be practically significant for real-time power-system operations: self-supervised training avoids expensive optimal labels, inference is orders of magnitude faster than state-of-the-art solvers, and the repair/completion layers are intended to make feasibility an architectural property. The derivations shown are largely sound: the balance repair layer in Eq. (5) is a differentiable feasible-point construction, the dual completion layer Clp produces dual-feasible points by construction, and the ALM-style primal-dual updates in Eqs. (12)-(13) are coherent. The experimental comparisons against Gurobi, Mosek, and CCGA on standard benchmarks are concrete and reproducible in spirit. However, the 'trustworthy by design' claim is not fully supported: the flagship PDL-SCOPF application does not have a by-construction guarantee for the contingency power-balance constraints, and the reported optimality gaps are empirical rather than certified. The contribution is therefore valuable as a synthesis and demonstration, but its stated guarantees need to be qualified.

major comments (3)
  1. [§7.3, Algorithm 2] The abstract and Section 8 state that optimization proxies are 'trustworthy by design' because they compute feasible solutions. For PDL-SCOPF, the text admits that Algorithm 2 'may not always satisfy the power balance constraints in the contingencies' and that these constraints are satisfied only 'when training completes in the experiments.' This makes feasibility an empirical property of a converged training run, not an architectural guarantee. Since PDL-SCOPF is one of the two headline applications and underlies the four-orders-of-magnitude speedup claim in Section 7.4, the central claim is overstated. The paper should either soften the 'by design' language to 'empirically feasible on the tested instances' or add a certified repair step for the contingency power-balance constraints.
  2. [§5.4.3, §6, §7.4, Table 13] The paper claims that optimization proxies 'provide quality guarantees on returned solutions.' The dual completion layer in Section 6 does construct dual-feasible points, hence valid lower bounds, by construction. However, the optimality gaps reported for the primal proxies in Tables 2 and 13 are empirical means over test sets, and Table 7 reports empirical dual-gap ratios; no certified primal-dual gap is produced for the applications. The phrase 'quality guarantees' should therefore be replaced by 'empirical quality' for primal solutions, or the paper should demonstrate a principled primal-dual gap procedure that ties the dual bound to the returned primal solution.
  3. [§8] Section 8 lists 'understanding how to derive effective repair layers for a wide range of applications' as an open problem. The general methodology therefore depends on an application-specific design step whose tractability is not established. This limitation is acknowledged in the conclusion, but it should also be reflected in the abstract: the 'by design' guarantee is conditional on the existence of an efficient repair or completion layer for the problem class. Stating this as a formal assumption at the outset, rather than as an open problem at the end, would make the scope of the claim precise.
minor comments (4)
  1. [§4.2] The subscript in the multiplier update contains a spurious '<' character: 'Mθt<+1(xi)' should be 'Mθ_{t+1}(xi)'.
  2. [§5.2, §7.4, §8] There are several typographical errors, including 'witout', 'imlpicit', 'opimization', 'PDF-SCOPF' for PDL-SCOPF, and 'strenghts'; a careful proofread is needed.
  3. [§7.3] The sentence 'As shown in Figure 2' in the description of the PDL-SCOPF primal network is misleading; the intended reference is likely Figure 10.
  4. [§7.3] The statement that Algorithm 2 may fail 'for reasons described in [37]' leaves a load-bearing explanation in a cited paper; a one-sentence summary of those reasons would make the limitation self-contained.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central repair/completion derivations are self-contained and validated against external solvers; self-citations are review-style, not load-bearing.

full rationale

The paper's core methodological derivations are self-contained and do not reduce to their inputs. In Section 5.2, the dedicated repair layers for the economic dispatch problem are constructed directly from the power balance and reserve constraints: the scaling formulas (5) and (6) produce a feasible point by construction whenever one exists, and the layer is not fitted to the target outputs. Section 6 derives the dual completion layer Clp(qz) = (qz, |cx - qz Ax|+, |cx - qz Ax|-) from the dual feasibility equation, so the claimed lower bound follows from standard LP duality rather than from any prediction being renamed as a guarantee. Section 7's primal-dual learning equations (12)-(15) are exactly the augmented Lagrangian method's primal and dual updates, and the method is trained self-supervised against the original objective, not against labels that encode the conclusion. The empirical claims are checked against external solvers (Gurobi, Mosek, CCGA) on PGLib benchmarks with held-out test sets, and the self-supervised variants do not use optimal solutions as training labels. The heavy self-citation to [9, 23, 36, 37] is normal for a review of the author's own research line, but the load-bearing derivations are restated in the text, so the citations are not carrying the logical argument. The Section 7.3 admission that the binary-search layer 'may not always satisfy the power balance constraints in the contingencies' weakens the 'trustworthy by design' wording for PDL-SCOPF, but that is a correctness or overclaim concern, not a circularity: the statement does not make the paper's conclusion equivalent to its assumptions by definition. Overall, no circular step is present.

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

The ledger is dominated by hand-chosen data-generation parameters and distributional assumptions. The methods themselves introduce no new physical or mathematical entities; the repair and completion layers are deterministic functions of the problem data. The PDL convergence assumption and the fixed-distribution assumption are load-bearing for the empirical claims.

free parameters (5)
  • Load scaling factor gamma = sampled U[0.8, 1.2]
    Data generation parameter chosen by hand for the ED/DCOPF experiments (Section 5.4.1); defines the input distribution and hence the reported gaps.
  • Nodal white noise eta = log-normal with mean 1, std 5%
    Chosen by hand to create training and test instances from PGLib benchmarks (Section 5.4.1).
  • Reserve requirement R = sampled uniformly between 100% and 200% of largest generator
    PGLib lacks reserve data, so this distribution is invented to mimic industry practice (Section 5.4.1); it controls the difficulty of the ED instances.
  • Reserve capacity fraction alpha_r = 5 * ||pbar||_inf / ||pbar||_1
    Ad hoc formula to set generator reserve capacities (Section 5.4.1); varies with system size and affects the feasibility landscape.
  • PDL hyperparameters (rho, alpha, tau, rho_max, T) = not specified numerically
    Algorithm 1 depends on the penalty coefficient, its update multiplier, violation tolerance, upper bound, and outer iterations. Numeric values are not given in this paper, so the SCOPF results are not fully reproducible from this preprint alone.
assumptions (5)
  • standard math The implicit function theorem and KKT conditions apply to differentiable repair layers
    Used in Section 5.1 to differentiate through the projection repair layer (4).
  • standard math Deep neural networks are universal approximators
    Invoked in Section 4.1 to justify learning the input/output mapping of an optimization problem.
  • domain assumption The augmented Lagrangian method converges for the convex problems considered
    Primal-dual learning in Section 7 mimics ALM; convergence of the learned networks is expected, not proven.
  • domain assumption The instance distribution P is known and stationary, and test instances come from the same distribution
    Stated in Section 2 and used in all experiments; distribution shift is not addressed and would break the empirical guarantees.
  • domain assumption PTDF and LODF linear models accurately represent transmission constraints
    Used in the ED and SCOPF models (Sections 3 and 7.1); standard in industry but an approximation of AC power flow.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Optimization Learning." pith.science (2026). https://pith.science/paper/URWLF3GM

@misc{pith2026250103443,
  author       = {Pith},
  title        = {Pith review of: Optimization Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/URWLF3GM}},
  note         = {Machine review of arXiv:2501.03443}
}
read the original abstract

This article introduces the concept of optimization learning, a methodology to design optimization proxies that learn the input/output mapping of parametric optimization problems. These optimization proxies are trustworthy by design: they compute feasible solutions to the underlying optimization problems, provide quality guarantees on the returned solutions, and scale to large instances. Optimization proxies are differentiable programs that combine traditional deep learning technology with repair or completion layers to produce feasible solutions. The article shows that optimization proxies can be trained end-to-end in a self-supervised way. It presents methodologies to provide performance guarantees and to scale optimization proxies to large-scale optimization problems. The potential of optimization proxies is highlighted through applications in power systems and, in particular, real-time risk assessment and security-constrained optimal power flow.

Figures

Figures reproduced from arXiv: 2501.03443 by the authors.

Figure 1
Figure 1. The machine learning training consists of series of forward and backward passes. At iteration [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 1
Figure 1. The Training of Supervised Learning Models. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The Architecture of Optimization Proxies. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figures from the paper (8 more)
Figure 3
Figure 3. Figure 3: End to End Training of Optimization Proxies. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png]
Figure 4
Figure 4. Figure 4: Explicit versus Implicit Layers in Deep Learning. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: The Optimization Proxy for the Economic Dispatch Problem (adapted from [9]). [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: A Risk Assessment Framework with Optimization Proxies (from [11]). [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: Probability of an Adverse Event with a Proxy-based Risk Assessment (from [11]). [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: The Structure of Dual Optimization Proxies. [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 9
Figure 9. Figure 9: The Architecture of Primal-Dual Learning (from [36]). [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: The Primal and Dual Networks of PDL-SCOPF (Adapted from [37]. [PITH_FULL_IMAGE:figures/full_fig_p018_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 44 canonical work pages

  1. [1]

    Agrawal, Amos B., S

    A. Agrawal, Amos B., S. Barratt, Boyd S., S. Diamond, and Z. Kolter. Differentiable Convex Optimization Layers. In Proceedings of the Neural Information Processing Systems Conference (NeurIPS), 2019. 22

    work page 2019

  2. [2]

    Optnet: Differentiable optimization as a layer in neural networks

    Brandon Amos and J Zico Kolter. Optnet: Differentiable optimization as a layer in neural networks. In ICML, pages 136–145. JMLR. org, 2017

    work page 2017

  3. [3]

    Recent develop- ments in security-constrained AC optimal power flow: Overview of challenge 1 in the ARPA-E Grid Optimization Competition

    Ignacio Aravena, Daniel K Molzahn, Shixuan Zhang, Cosmin G Petra, Frank E Curtis, Shenyiny- ing Tu, Andreas W¨ achter, Ermin Wei, Elizabeth Wong, Amin Gholami, et al. Recent develop- ments in security-constrained AC optimal power flow: Overview of challenge 1 in the ARPA-E Grid Optimization Competition. arXiv preprint arXiv:2206.07843 , 2022

    arXiv 2022

  4. [4]

    Layer normalization

    Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E Hinton. Layer normalization. arXiv preprint arXiv:1607.06450, 2016

    arXiv 2016

  5. [5]

    The power grid library for benchmarking AC optimal power flow algorithms

    Sogol Babaeinejadsarookolaee, Adam Birchfield, Richard D Christie, Carleton Coffrin, Christo- pher DeMarco, Ruisheng Diao, Michael Ferris, Stephane Fliscounakis, Scott Greene, Renke Huang, et al. The power grid library for benchmarking AC optimal power flow algorithms. arXiv preprint arXiv:1908.02788 , 2019

    arXiv 1908

  6. [6]

    Sogol Babaeinejadsarookolaee, Adam Birchfield, Richard D. Christie, Carleton Coffrin, Christo- pher DeMarco, Ruisheng Diao, Michael Ferris, Stephane Fliscounakis, Scott Greene, Renke Huang, Cedric Josz, Roman Korab, Bernard Lesieutre, Jean Maeght, Terrence W. K. Mak, Daniel K. Molzahn, Thomas J. Overbye, Patrick Panciatici, Byungkwon Park, Jonathan Snod- ...

    work page 2019

  7. [7]

    Machine learning for combinatorial optimiza- tion: A methodological tour d’horizon

    Yoshua Bengio, Andrea Lodi, and Antoine Prouvost. Machine learning for combinatorial optimiza- tion: A methodological tour d’horizon. European Journal of Operational Research , 290(2):405– 421, 2021

    work page 2021

  8. [8]

    Two-Stage Learning For the Flexible Job Shop Scheduling Problem

    Wenbo Chen, Reem Khir, and Pascal Van Hentenryck. Two-stage learning for the flexible job shop scheduling problem. CoRR, abs/2301.09703, 2023

    work page Pith review arXiv 2023

Show all 49 references
  1. [9]

    End-to-end feasible optimization proxies for large-scale economic dispatch

    Wenbo Chen, Mathieu Tanneau, and Pascal Van Hentenryck. End-to-end feasible optimization proxies for large-scale economic dispatch. IEEE Transactions on Power Systems (to appear) , 2023

    2023

  2. [10]

    End-to-End Feasible Optimization Proxies for Large-Scale Economic Dispatch

    Wenbo Chen, Mathieu Tanneau, and Pascal Van Hentenryck. End-to-End Feasible Optimization Proxies for Large-Scale Economic Dispatch. arXiv (preprint), 2023

    2023

  3. [11]

    Real-time risk analysis with opti- mization proxies

    Wenbo Chen, Mathieu Tanneau, and Pascal Van Hentenryck. Real-time risk analysis with opti- mization proxies. In Proceedings of the 23rd International Power Systems Computation Confer- ence (PSCC’2024), Paris-Saclay, France, June 2024

    2024

  4. [12]

    Donti, David Rolnick, and J Zico Kolter

    Priya L. Donti, David Rolnick, and J Zico Kolter. DC3: A learning method for optimization with hard constraints. In International Conference on Learning Representations , 2021

    2021

  5. [13]

    Optimizing primary response in preventive security-constrained optimal power flow.IEEE Systems Journal, 12(1):414– 423, 2016

    Yury Dvorkin, Pierre Henneaux, Daniel S Kirschen, and Hrvoje Pandˇ zi´ c. Optimizing primary response in preventive security-constrained optimal power flow.IEEE Systems Journal, 12(1):414– 423, 2016

    2016

  6. [14]

    Fioretto Ferdinando, Terrence W. K. Mak, and Pascal Van Hentenryck. Predicting AC optimal power flows: Combining deep learning and lagrangian dual methods. In The Thirty-Fourth AAAI Conference on Artificial Intelligence, AAAI , pages 630–637. AAAI Press, February 2020

    2020

  7. [15]

    Ferdinando Fioretto, Pascal Van Hentenryck, Terrence W. K. Mak, Cuong Tran, Federico Baldo, and Michele Lombardi. Lagrangian duality for constrained deep learning. In Yuxiao Dong, Geor- giana Ifrim, Dunja Mladeni´ c, Craig Saunders, and Sofie Van Hoecke, editors, Machine Learn...

    2021

  8. [16]

    Gould, R

    S. Gould, R. Hartley, and D. Campbell. Deep declarative networks. IEEE Transactions on Pattern Analysis and Machine Intelligence , 44(08):3988–4004, aug 2022

    2022

  9. [17]

    “Grid optimization competition datasets, 2018

    Grid Optimization Competition. “Grid optimization competition datasets, 2018

    2018

  10. [18]

    Direct calculation of line outage distribution factors

    Jiachun Guo, Yong Fu, Zuyi Li, and Mohammad Shahidehpour. Direct calculation of line outage distribution factors. IEEE Transactions on Power Systems , 24(3):1633–1634, 2009

    2009

  11. [19]

    Fast Simultaneous Feasibility Test for Security Constrained Unit Commitment, 2022

    Jesse Holzer, Yonghong Chen, Zhongyu Wu, Feng Pan, and Arun Veeramany. Fast Simultaneous Feasibility Test for Security Constrained Unit Commitment, 2022

    2022

  12. [20]

    Multilayer feedforward networks are universal approximators

    Kurt Hornik, Maxwell Stinchcombe, and Halbert White. Multilayer feedforward networks are universal approximators. Neural Networks, 2(5):359–366, 1989

    1989

  13. [21]

    DeepOPF-NGT: Fast No Ground Truth Deep Learning- Based Approach for AC-OPF Problems

    Wanjun Huang and Minghua Chen. DeepOPF-NGT: Fast No Ground Truth Deep Learning- Based Approach for AC-OPF Problems. In ICML 2021 Workshop Tackling Climate Change with Machine Learning, 2021

    2021

  14. [22]

    AC Power Flow Data in MATPOWER and QCQP Format: iTesla, RTE Snapshots, and PEGASE, 2016

    C´ edric Josz, St´ ephane Fliscounakis, Jean Maeght, and Patrick Panciatici. AC Power Flow Data in MATPOWER and QCQP Format: iTesla, RTE Snapshots, and PEGASE, 2016

    2016

  15. [23]

    Dual interior-point optimiza- tion learning, 2024

    Michael Klamkin, Mathieu Tanneau, and Pascal Van Hentenryck. Dual interior-point optimiza- tion learning, 2024

    2024

  16. [24]

    A new computationally simple approach for implement- ing neural networks with output hard constraints

    Andrei V Konstantinov and Lev V Utkin. A new computationally simple approach for implement- ing neural networks with output hard constraints. In Doklady Mathematics, pages 1–9. Springer, 2024

    2024

  17. [25]

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

    James Kotary, Ferdinando Fioretto, and Pascal Van Hentenryck. Fast approximations for job shop scheduling: A lagrangian dual deep learning method. In In the Thirty-Sixth AAAI Conference on Artificial Intelligence (AAAI-22) , February 2022

    2022

  18. [26]

    End-to-end con- strained optimization learning: A survey

    James Kotary, Ferdinando Fioretto, Pascal Van Hentenryck, and Bryan Wilder. End-to-end con- strained optimization learning: A survey. In Zhi-Hua Zhou, editor, Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, IJCAI-21 , pages 4475–4482. In...

    2021

  19. [27]

    The Implicit Function Theorem

    Steven Krantz and Harold Parks. The Implicit Function Theorem . Modern Birkhauser Classics. Birkhauser. ISBN 0-8176-4285-4, 2003

    2003

  20. [28]

    Learning to Solve Optimization Problems With Hard Linear Constraints

    Meiyi Li, Soheil Kolouri, and Javad Mohammadi. Learning to Solve Optimization Problems With Hard Linear Constraints. IEEE Access, 11:59995–60004, 2023

    2023

  21. [29]

    The security-constrained commitment and dispatch for midwest iso day-ahead co-optimized energy and ancillary service market

    Xingwang Ma, Haili Song, Mingguo Hong, Jie Wan, Yonghong Chen, and Eugene Zak. The security-constrained commitment and dispatch for midwest iso day-ahead co-optimized energy and ancillary service market. In 2009 IEEE Power & Energy Society General Meeting , pages 1–8, 2009

    2009

  22. [30]

    Schedule 28A – Demand Curves for Transmission Constraints, 2019

    MISO. Schedule 28A – Demand Curves for Transmission Constraints, 2019

    2019

  23. [31]

    Real-time energy and operating reserve market software formulations and business logic,

    MISO. Real-time energy and operating reserve market software formulations and business logic,

  24. [32]

    Schedule 28 – Demand Curves for TOperating Reserves, 2023

    MISO. Schedule 28 – Demand Curves for TOperating Reserves, 2023

    2023

  25. [33]

    MOSEK Optimizer API for Julia 10.1.24 , 2022

    MOSEK ApS. MOSEK Optimizer API for Julia 10.1.24 , 2022

    2022

  26. [34]

    Optimization-based learning for dynamic load planning in trucking service networks, 2024

    Ritesh Ojha, Wenbo Chen, Hanyu Zhang, Reem Khir, Alan Erera, and Pascal Van Hentenryck. Optimization-based learning for dynamic load planning in trucking service networks, 2024. 24

    2024

  27. [35]

    DeepOPF: A Deep Neural Network Approach for Security-Constrained DC OPF

    Xiang Pan, Tianyu Zhao, Minghua Chen, and Shengyu Zhang. DeepOPF: A Deep Neural Network Approach for Security-Constrained DC OPF. IEEE Transactions on Power Systems , 36(3):1725– 1735, 2020

    2020

  28. [36]

    Self-supervised primal-dual learning for constrained optimization

    Seonho Park and Pascal Van Hentenryck. Self-supervised primal-dual learning for constrained optimization. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 37, pages 4052–4060, 2023

    2023

  29. [37]

    Self-supervised learning for large-scale preventive security constrained dc optimal power flow

    Seonho Park and Pascal Van Hentenryck. Self-supervised learning for large-scale preventive security constrained dc optimal power flow. IEEE Transactions on Power Systems , pages 1–11, 2024

    2024

  30. [38]

    Dual Conic Proxies for AC Optimal Power Flow

    Guancheng Qiu, Mathieu Tanneau, and Pascal Van Hentenryck. Dual Conic Proxies for AC Optimal Power Flow. In Proceedings of the 23rd International Power Systems Computation Conference (PSCC’2024), Paris-Saclay, France, June 2024

    2024

  31. [39]

    Dual lagrangian learning for conic optimization, 2024

    Mathieu Tanneau and Pascal Van Hentenryck. Dual lagrangian learning for conic optimization, 2024

    2024

  32. [40]

    Security constrained unit commitment using line outage distribution factors

    Diego A Tejada-Arango, Pedro S´ anchez-Martın, and Andres Ramos. Security constrained unit commitment using line outage distribution factors. IEEE Transactions on power systems , 33(1):329–337, 2017

    2017

  33. [41]

    Rayen: Imposition of hard convex con- straints on neural networks

    Jesus Tordesillas, Jonathan P How, and Marco Hutter. Rayen: Imposition of hard convex con- straints on neural networks. 2023

    2023

  34. [42]

    of Electrical Engineering

    University of Washington, Dept. of Electrical Engineering. Power systems test case archive, 1999

    1999

  35. [43]

    Machine Learning for Optimal Power Flows

    Pascal Van Hentenryck. Machine Learning for Optimal Power Flows. Tutorials in Operations Research, pages 62–82, December 2021

    2021

  36. [44]

    Combining deep learning and optimization for preventive security-constrained DC optimal power flow

    Alexandre Velloso and Pascal Van Hentenryck. Combining deep learning and optimization for preventive security-constrained DC optimal power flow. IEEE Transactions on Power Systems , 36(4):3618–3628, 2021

    2021

  37. [45]

    An exact and scalable problem decomposition for security-constrained optimal power flow

    Alexandre Velloso, Pascal Van Hentenryck, and Emma S Johnson. An exact and scalable problem decomposition for security-constrained optimal power flow. Electric Power Systems Research , 195:106677, 2021

    2021

  38. [46]

    Fast optimal power flow with guarantees via an unsu- pervised generative model

    Junfei Wang and Pirathayini Srikantha. Fast optimal power flow with guarantees via an unsu- pervised generative model. IEEE Transactions on Power Systems , 2022

    2022

  39. [47]

    Melding the data-decisions pipeline: Decision- focused learning for combinatorial optimization

    Bryan Wilder, Bistra Dilkina, and Milind Tambe. Melding the data-decisions pipeline: Decision- focused learning for combinatorial optimization. In AAAI, volume 33, pages 1658–1665, 2019

    2019

  40. [48]

    Reinforcement learning from optimiza- tion proxy for ride-hailing vehicle relocation

    Enpeng Yuan, Wenbo Chen, and Pascal Van Hentenryck. Reinforcement learning from optimiza- tion proxy for ride-hailing vehicle relocation. J. Artif. Intell. Res. , 75:985–1002, 2022. 25

    2022

  41. [2022]

    Business Practices Manual Energy and Operating Reserve Markets Attachment D

Pith tools

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