Pith. sign in

REVIEW 4 major objections 3 minor 50 references

Sobolev Training of End-to-End Optimization Proxies

T0 review · 4 major / 3 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Adding solver-sensitivity matching to proxy training yields quadratic error bounds and cuts large-scale AC-OPF error by up to 56%.

desk verdict A useful masking recipe for Sobolev-trained optimization proxies, with an elegant but disconnected theory; worth refereeing, but the guarantee does not cover the actual method. read the letter →

arxiv 2505.11342 v1 pith:TVQM32XL submitted 2025-05-16 cs.LG math.OC

Andrew W. Rosemberg , Joaquim Dias Garcia , Russell Bent , Pascal Van Hentenryck This is my paper · ORCID
classification cs.LGmath.OC MSC 68T0790C3190C26
keywords SobolevtrainingoptimizationproxiesparametricsensitivityanalysisACoptimalpowerflowmean-varianceportfolioJacobianmatchingself-supervisedlearning
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 an optimization proxy—a neural network trained to imitate the solution map of a parametric optimization problem—improves markedly when its loss also matches the solver's local derivatives, not just its outputs. The paper proves that if the proxy reproduces both the value and the derivative matrix of the true solution map at every training point, and both maps have Lipschitz derivatives, the worst-case approximation error is at most a constant times the square of the training-set covering radius, a quadratic improvement over value-only matching. On three large alternating-current optimal power flow (AC-OPF) benchmarks, the supervised derivative-matching proxy cuts mean squared error by up to 56% and reduces median worst-case constraint violation by up to a factor of four while keeping the optimality gap below 0.22%. On a self-supervised mean-variance portfolio task, the same derivative signal halves the average optimality gap in the medium-risk region and matches the baseline in the tight-risk region, which matters because a single forward pass could then replace an iterative solver in time-critical or safety-critical decisions.

What carries the argument

The key object is the masked Sobolev loss, $L(\theta)=\frac{1}{N}\sum_i \ell(\hat{g}_\theta(p_i),g(p_i)) + \frac{\lambda}{N}\sum_i \ell_d(M\odot D\hat{g}_\theta(p_i), M\odot Dg(p_i))$, which combines value regression with a penalty on selected derivative entries. The proof engine is a first-order Taylor expansion with integral remainder: exact value and Jacobian interpolation at a training point cancels the constant and linear terms in $\hat{g}_\theta-g$, leaving only second-order remainders bounded by $\frac{1}{2}(M_g+M_{\hat{g}})\|h\|^2$, which yields the covering-radius-squared guarantee.

What would settle it

On a problem with an analytically known solution map $g$ and known derivative-Lipschitz constant $M_g$, train a smooth network with exact value and dense Jacobian interpolation on training grids of two different covering radii $\delta$; if the measured uniform error does not scale like $\delta^2$, or exceeds $\frac{1}{2}(M_g+M_{\hat{g}})\delta^2$, then Theorem 3 is not the mechanism behind the reported gains.

Watch

Extended reading notes

Core claim

The central claim is that solver sensitivities are a cheap, high-value training signal for end-to-end optimization proxies. Theorem 3 shows that matching both the solution value and its first-order derivative at every training point yields $\sup_{p\in P}\|\hat{g}_\theta(p)-g(p)\| \le \frac{1}{2}(M_g+M_{\hat{g}})\delta^2$, provided the true solution map and the network are $C^1$ with Lipschitz Jacobians on the compact parameter domain. In practice the loss matches only a randomly masked subset of Jacobian entries, and the paper argues that extreme sparsity stabilizes training and improves test error. Empirically, supervised derivative-matching training on AC-OPF is reported to cut mean squared error by up to 56% and median worst-case infeasibility by up to 400%, while the self-supervised portfolio experiment reports the average optimality gap in the medium-risk region dropping from 18.9% to 8.7%.

Load-bearing premise

All theoretical guarantees rest on the premise that at every training point the proxy exactly reproduces both the solver's output and its full derivative matrix, using smooth activations with bounded curvature and a parametric problem regular enough for the solution map to be smooth; the paper's own experiments instead keep only 5-25% of Jacobian entries, use ReLU or LeakyReLU activations, and note that regularity conditions occasionally fail in AC-OPF, so the stated bound does not directly govern the reported models.

Editorial extensions

If this is right

  • Supervised derivative matching changes the accuracy-feasibility trade-off for AC-OPF proxies: it lowers mean squared error and sharply reduces worst-case violations while keeping optimality gaps under 0.22%.
  • Self-supervised derivative matching offers a path to train proxies without optimal labels; in the medium-risk portfolio region it halves the average optimality gap.
  • Matching derivatives reduces the sample complexity of learning the solution map, since the theoretical error depends on the square of the training-set radius rather than the radius itself.
  • Randomly masking most Jacobian entries is not only a memory saver; the paper's ablations indicate that dense Jacobian supervision actively hurts performance.

Reading between the lines

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

  • A direct testable extension is to embed the derivative-matching proxy in an outer optimization loop and measure convergence speed; the paper argues for gradient fidelity but never measures its downstream effect.
  • The sharp split between tight-risk and medium-risk portfolio regimes suggests training a gated model that delegates to the value-only proxy when the risk budget is below 10% of capital and to the derivative-matching proxy otherwise; the paper names this mixture-of-experts idea but does not implement it.
  • The quadratic error bound implies derivative labels could substitute for denser parameter sampling; a grid-refinement experiment comparing value-only and derivative-matching training at matched sample counts would quantify that trade-off.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 3 minor

Summary. The paper proposes a masked Sobolev training objective for optimization proxies, in which a standard value-regression loss is augmented by a sparsified Jacobian-matching term that aligns the proxy's derivatives with solver sensitivities. The authors claim uniform approximation guarantees, including an O(delta^2) bound under joint value and Jacobian interpolation, and report empirical results on three PGLib AC-OPF benchmarks plus a mean-variance portfolio task. In the supervised AC-OPF experiments, the Sobolev-trained proxy generally improves MSE and constraint-violation metrics over an MSE-only baseline; in the semi-self-supervised portfolio task, it improves average optimality gap in the medium-risk regime but not in the tight-risk regime.

Significance. If Theorem 3 applied to the actual trained models, the paper would make a substantial contribution by connecting derivative matching to a uniform approximation guarantee with a quadratic dependence on training-set density. The proof of Theorem 3 is short and mathematically correct under its stated assumptions, and the reported empirical gains on large PGLib instances are suggestive. However, the central theoretical guarantee is not connected to the method as implemented: the paper's own loss is finite-sample and mask-sparsified, the activations used in most experiments are not C^2, and the AC-OPF solution map is not globally C^1. These gaps mean the headline theoretical result currently serves mainly as motivation rather than as a guarantee for the evaluated algorithm.

major comments (4)
  1. [Section 4, Theorem 3] The theorem assumes that Sobolev training enforces exact interpolation of both g(p_i) and Dg(p_i) at every training point, as stated in the paragraph before Assumption 1. The loss defined in Section 3 is a finite-sample weighted sum of a value term and a randomly masked Jacobian term, and no mechanism in the training procedure enforces exact interpolation. With the reported masks keeping only 5-25% of Jacobian entries, D\hat{g}_theta(p_i) is not equal to Dg(p_i) for most entries, so Theorem 3 cannot be invoked for the models whose results are reported. Please either train with an interpolation-enforcing mechanism, extend the bound to the masked finite-sample objective with an explicit residual term depending on the training loss, or state clearly that Theorem 3 is an idealized result that does not govern the experiments.
  2. [Remark 1 and Table 3] Remark 1 restricts Theorems 1-3 to proxies with C^2 activations of bounded curvature, giving tanh and softplus as examples. Table 3 lists ReLU activations for PEGASE-1k and RTE-6k and LeakyReLU for the Markowitz model; these are not C^2, so the theoretical guarantee does not cover them. The sigmoid model on IEEE-300 is C^2, but it still does not satisfy the full-Jacobian interpolation assumption because of the masked Jacobian loss. Consequently, none of the four experimental configurations satisfies the assumptions needed for Theorem 3 as stated.
  3. [Assumption 2 and Section 7] Theorem 3 requires g to be C^1 on the entire compact parameter set P, with a Lipschitz Jacobian. AC-OPF solution maps are only piecewise smooth because the set of active constraints can change as demand parameters vary, and Section 7 concedes that 'irregular cases were present' in the experiments. The theorem's conclusion is a global supremum bound, so it cannot tolerate nonsmooth regions even if such regions are rare. The statement that irregular cases 'did not materially affect overall proxy performance' is an empirical claim and does not repair the theoretical gap; the theory should either restrict P to a smooth region or be reformulated for piecewise-smooth maps.
  4. [Section 6] The portfolio study is described as self-supervised and the introduction to Section 6 states that SSL 'avoids the need for optimal solutions.' However, the Sobolev variant is trained on 'solver sensitivities,' and Equation (1) shows that these sensitivities are obtained by differentiating the KKT system at an optimal primal-dual solution. Computing Dg(p_i) therefore requires computing x*(p_i) and lambda*(p_i), so the method does not in fact avoid the need for optimal solutions. The paper should clarify what 'semi' self-supervised means, or explain how sensitivities can be produced without solving the optimization problem.
minor comments (3)
  1. [Table 3] The 'Mask sparsity (%)' column lists the values 0.145, 0.305, 0.125, and 0.355, which are inconsistent with the usual meaning of a percentage and with the paper's statement that 5-25% of Jacobian entries are retained. Please clarify whether these numbers are retained fractions or sparsity percentages, and ensure they match the values used in Appendix D.
  2. [Section 5.2] The definition of RMI_i normalizes by max_{i,j}(Infeas^{Sobolev}_{i,j}), the largest violation of the Sobolev model only, rather than by a maximum over both models. This makes the reported relative improvements hard to interpret and highly sensitive to the scale of Sobolev violations; normalizing by the maximum over both baselines would be more natural.
  3. [Throughout] There are several typographical issues: 'intwo' in the abstract, 'refereed to' in Section 5.1, and 'Xases' in the caption of Figure 2(b). Please proofread the manuscript.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Theorem 3 is a conditional interpolation bound with an explicit proof; empirical gains are externally benchmarked, and the admitted theory–implementation gap is an applicability issue, not a circular one.

full rationale

The central theoretical result, Theorem 3, is a conditional statement: if the proxy exactly interpolates both g and Dg at every training point and both maps are derivative-Lipschitz, then the uniform error is at most 1/2(Mg + M_hatg) delta^2. The proof uses the fundamental theorem of calculus and the triangle inequality; the bound follows from the assumptions rather than being introduced as a conclusion or fitted parameter. There is no self-definitional step where X is defined in terms of Y, and no fitted input is relabeled as a prediction. The empirical claims are evaluated on held-out PGLib AC-OPF test instances against an MSE-trained baseline, with solver labels and sensitivities obtained from external tools (IPOPT, DiffOpt.jl), not from the proxy itself; the self-citations to DiffOpt.jl, L2O.jl, and prior OPF papers are tooling and context, not load-bearing for the theoretical guarantee. The paper's own Section 7 explicitly concedes that random Jacobian masking (keeping only 5–25% of entries) and occasional degenerate sensitivities mean the exact-interpolation, full-Jacobian, C^2 setting of Theorem 3 does not strictly match the implemented models; Remark 1's claim that C^2 activations apply is inconsistent with the ReLU/LeakyReLU entries in Table 3. That is an applicability or correctness gap, not circularity, because the theorem does not assume its conclusion. No specific reduction of the claimed result to its inputs was found.

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

The theoretical claim is clean but relies on idealized interpolation and smoothness assumptions that the practical pipeline does not satisfy. The empirical pipeline has several tuned hyperparameters (lambda, mask density, penalty weights) that are not derived from theory.

free parameters (3)
  • loss trade-off weight lambda = 0.145, 0.305, 0.125, 0.355 per benchmark (Table 3)
    Balances value loss and derivative loss; chosen by hand and tuned per task. A free hyperparameter not fixed by theory.
  • Jacobian mask retained fraction = 5-25% kept, with densities per benchmark in Table 3
    The fraction of Jacobian entries supervised is a tuned hyperparameter; Appendix D shows mask density strongly affects test MSE.
  • self-supervised penalty weights beta, gamma = not reported numerically
    In Section 3, the self-supervised loss uses beta and gamma penalty weights; their values are not specified, a reproducibility gap.
assumptions (4)
  • domain assumption LICQ, SOSC, and strict complementarity hold at the sampled solutions so the KKT sensitivity map is differentiable and non-singular.
    Invoked in Section 2.1 to justify Eq. (1), and in Remark 1 to ensure g is C^{1,1}. The authors admit in Section 7 that occasional degeneracies occur.
  • domain assumption The solution map g and proxy hat g are C^1 with Lipschitz Jacobians on compact P (Assumptions 1 and 2).
    Hypotheses of Theorems 1-3. The proxy activations in the experiments (ReLU, LeakyReLU) are not C^2, so the hypothesis is not satisfied by the trained networks.
  • ad hoc to paper The proxy exactly interpolates g and Dg on the training set.
    Section 4 assumes exact equality at training points; the actual loss is a weighted finite-sample regression with masked Jacobians, so exact interpolation is not achieved.
  • standard math Parameter space P is compact with finite covering radius delta.
    Used to define delta and take suprema in Theorems 1-3; reasonable for bounded demand and covariance data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Sobolev Training of End-to-End Optimization Proxies." pith.science (2026). https://pith.science/paper/TVQM32XL

@misc{pith2026250511342,
  author       = {Pith},
  title        = {Pith review of: Sobolev Training of End-to-End Optimization Proxies},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TVQM32XL}},
  note         = {Machine review of arXiv:2505.11342}
}
read the original abstract

Optimization proxies - machine learning models trained to approximate the solution mapping of parametric optimization problems in a single forward pass - offer dramatic reductions in inference time compared to traditional iterative solvers. This work investigates the integration of solver sensitivities into such end to end proxies via a Sobolev training paradigm and does so in two distinct settings: (i) fully supervised proxies, where exact solver outputs and sensitivities are available, and (ii) self supervised proxies that rely only on the objective and constraint structure of the underlying optimization problem. By augmenting the standard training loss with directional derivative information extracted from the solver, the proxy aligns both its predicted solutions and local derivatives with those of the optimizer. Under Lipschitz continuity assumptions on the true solution mapping, matching first order sensitivities is shown to yield uniform approximation error proportional to the training set covering radius. Empirically, different impacts are observed in each studied setting. On three large Alternating Current Optimal Power Flow benchmarks, supervised Sobolev training cuts mean squared error by up to 56 percent and the median worst case constraint violation by up to 400 percent while keeping the optimality gap below 0.22 percent. For a mean variance portfolio task trained without labeled solutions, self supervised Sobolev training halves the average optimality gap in the medium risk region (standard deviation above 10 percent of budget) and matches the baseline elsewhere. Together, these results highlight Sobolev training whether supervised or self supervised as a path to fast reliable surrogates for safety critical large scale optimization workloads.

Figures

Figures reproduced from arXiv: 2505.11342 by the authors.

Figure 1
Figure 1. End-to-end pipeline for Sobolev training of optimization proxies. The [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Comparison of NN Sobolev vs Benchmark across the Three Test Xases. [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Change in Relative Optimality Gap ( [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Change in Relative Optimality Gap ( [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 35 canonical work pages

  1. [1]

    Optimization learning.arXiv preprint arXiv:2501.03443, 2025

    Pascal Van Hentenryck. Optimization learning.arXiv preprint arXiv:2501.03443, 2025

    arXiv 2025

  2. [2]

    Tutorial on amortized optimization.Foundations and Trends® in Machine Learning, 16(5):592–732, 2023

    Brandon Amos et al. Tutorial on amortized optimization.Foundations and Trends® in Machine Learning, 16(5):592–732, 2023

    2023

  3. [3]

    Dc3: A learning method for optimization with hard constraints.arXiv preprint arXiv:2104.12225, 2021

    Priya L Donti, David Rolnick, and J Zico Kolter. Dc3: A learning method for optimization with hard constraints.arXiv preprint arXiv:2104.12225, 2021

    arXiv 2021

  4. [4]

    Sobolev training for neural networks.Advances in neural information processing systems, 30, 2017

    Wojciech M Czarnecki, Simon Osindero, Max Jaderberg, Grzegorz Swirszcz, and Razvan Pascanu. Sobolev training for neural networks.Advances in neural information processing systems, 30, 2017

    work page 2017

  5. [5]

    Sobolev trained neural network surrogate models for optimization.Computers & Chemical Engineering, 153:107419, 2021

    Calvin Tsay. Sobolev trained neural network surrogate models for optimization.Computers & Chemical Engineering, 153:107419, 2021

    work page 2021

  6. [6]

    Sensitivity analysis for parametric nonlinear programming: A tutorial.arXiv preprint arXiv:2504.15851, 2025

    François Pacaud. Sensitivity analysis for parametric nonlinear programming: A tutorial.arXiv preprint arXiv:2504.15851, 2025

    arXiv 2025

  7. [8]

    Sensitivity analysis for nonlinear programming using penalty methods

    Anthony V Fiacco. Sensitivity analysis for nonlinear programming using penalty methods. Mathematical programming, 10(1):287–311, 1976

    work page 1976

  8. [9]

    Strongly stable stationary solutions in nonlinear programs.Analysis and Computation of Fixed Points, pages 93–138, 1980

    Masakazu Kojima. Strongly stable stationary solutions in nonlinear programs.Analysis and Computation of Fixed Points, pages 93–138, 1980

    work page 1980

Show all 50 references
  1. [10]

    Solution point differentiability without strict complementarity in nonlinear programming.Sensitivity, Stability and Parametric Analysis, pages 127–138, 1984

    Krisorn Jittorntrum. Solution point differentiability without strict complementarity in nonlinear programming.Sensitivity, Stability and Parametric Analysis, pages 127–138, 1984

    1984

  2. [11]

    Second order sensitivity analysis and asymptotic theory of parametrized nonlinear programs.Mathematical Programming, 33:280–299, 1985

    Alexander Shapiro. Second order sensitivity analysis and asymptotic theory of parametrized nonlinear programs.Mathematical Programming, 33:280–299, 1985

    1985

  3. [12]

    Directional derivatives of the solution of a parametric nonlinear program.Mathematical Programming, 70(1-3):159–172, 1995

    Daniel Ralph and Stephan Dempe. Directional derivatives of the solution of a parametric nonlinear program.Mathematical Programming, 70(1-3):159–172, 1995

    1995

  4. [13]

    Foundations of bilevel programming.Springer Science & Business Media, 2002

    Stephan Dempe. Foundations of bilevel programming.Springer Science & Business Media, 2002

    2002

  5. [14]

    On differential stability in stochastic programming.Mathematical Pro- gramming, 47(1):107–116, 1990

    Alexander Shapiro. On differential stability in stochastic programming.Mathematical Pro- gramming, 47(1):107–116, 1990

    1990

  6. [15]

    Asymptotic analysis of stochastic programs.Annals of Operations Research, 30:169–186, 1991

    Alexander Shapiro. Asymptotic analysis of stochastic programs.Annals of Operations Research, 30:169–186, 1991. 10

    1991

  7. [16]

    Zavala and Lorenz T

    Victor M. Zavala and Lorenz T. Biegler. The advanced-step nmpc controller: Optimality, stability and robustness.Automatica, 45(1):86–93, 2009

    2009

  8. [17]

    Fast economic model predictive control based on nlp-sensitivities.Journal of Process Control, 24(8):1260–1272, 2014

    Johannes Jäschke, Xue Yang, and Lorenz T Biegler. Fast economic model predictive control based on nlp-sensitivities.Journal of Process Control, 24(8):1260–1272, 2014

    2014

  9. [18]

    Optimal sensitivity based on ipopt

    Hans Pirnay, Rodrigo López-Negrete, and Lorenz T Biegler. Optimal sensitivity based on ipopt. Mathematical Programming Computation, 4:307–331, 2012

    2012

  10. [19]

    Casadi: a software framework for nonlinear optimization and optimal control.Mathematical Programming Computation, 11:1–36, 2019

    Joel AE Andersson, Joris Gillis, Greg Horn, James B Rawlings, and Moritz Diehl. Casadi: a software framework for nonlinear optimization and optimal control.Mathematical Programming Computation, 11:1–36, 2019

    2019

  11. [20]

    Parallelizing the dual revised simplex method.Mathematical Programming Computation, 10(1):119–142, 2018

    Qi Huangfu and JA Julian Hall. Parallelizing the dual revised simplex method.Mathematical Programming Computation, 10(1):119–142, 2018

    2018

  12. [21]

    Accelerating optimal power flow with GPUs: SIMD abstraction of nonlinear programs and condensed-space interior-point methods

    Sungho Shin, François Pacaud, and Mihai Anitescu. Accelerating optimal power flow with GPUs: SIMD abstraction of nonlinear programs and condensed-space interior-point methods. arXiv preprint arXiv:2307.16830, 2023

    2023 arXiv

  13. [22]

    Jump 1.0: Recent improvements to a modeling language for mathematical optimization

    Miles Lubin, Oscar Dowson, Joaquim Dias Garcia, Joey Huchette, Benoît Legat, and Juan Pablo Vielma. Jump 1.0: Recent improvements to a modeling language for mathematical optimization. Mathematical Programming Computation, 15(3):581–589, 2023

    2023

  14. [23]

    A differentiable programming system to bridge machine learning and scientific computing.arXiv preprint arXiv:1907.07587, 2019

    Mike Innes, Alan Edelman, Keno Fischer, Chris Rackauckas, Elliot Saba, Viral B Shah, and Will Tebbutt. A differentiable programming system to bridge machine learning and scientific computing.arXiv preprint arXiv:1907.07587, 2019

    1907 arXiv

  15. [24]

    Optnet: Differentiable optimization as a layer in neural networks

    Brandon Amos and J Zico Kolter. Optnet: Differentiable optimization as a layer in neural networks. InInternational conference on machine learning, pages 136–145. PMLR, 2017

    2017

  16. [25]

    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

  17. [26]

    Differentiable convex optimization layers.Advances in neural information processing systems, 32, 2019

    Akshay Agrawal, Brandon Amos, Shane Barratt, Stephen Boyd, Steven Diamond, and J Zico Kolter. Differentiable convex optimization layers.Advances in neural information processing systems, 32, 2019

    2019

  18. [27]

    Theseus: A Library for Differentiable Nonlinear Optimization

    Luis Pineda, Taosha Fan, Maurizio Monge, Shobha Venkataraman, Paloma Sodhi, Ricky TQ Chen, Joseph Ortiz, Daniel DeTone, Austin Wang, Stuart Anderson, Jing Dong, Brandon Amos, and Mustafa Mukadam. Theseus: A Library for Differentiable Nonlinear Optimization. Advances in Neural ...

    2022

  19. [28]

    Flexible differen- tiable optimization via model transformations.INFORMS Journal on Computing, 36(2):456– 478, 2024

    Mathieu Besançon, Joaquim Dias Garcia, Benoît Legat, and Akshay Sharma. Flexible differen- tiable optimization via model transformations.INFORMS Journal on Computing, 36(2):456– 478, 2024

    2024

  20. [29]

    Config: Towards conflict-free training of physics informed neural networks.arXiv preprint arXiv:2408.11104, 2024

    Qiang Liu, Mengyu Chu, and Nils Thuerey. Config: Towards conflict-free training of physics informed neural networks.arXiv preprint arXiv:2408.11104, 2024

    2024 arXiv

  21. [30]

    Large-scale nonlinear programming using Ipopt: An integrating framework for enterprise-wide dynamic optimization.Computers & Chemical Engineering, 33(3):575–582, 2009

    Lorenz T Biegler and Victor M Zavala. Large-scale nonlinear programming using Ipopt: An integrating framework for enterprise-wide dynamic optimization.Computers & Chemical Engineering, 33(3):575–582, 2009

    2009

  22. [31]

    A convex neural network solver for dcopf with generalization guarantees.IEEE Transactions on Control of Network Systems, 9(2):719–730, 2022

    Ling Zhang, Yize Chen, and Baosen Zhang. A convex neural network solver for dcopf with generalization guarantees.IEEE Transactions on Control of Network Systems, 9(2):719–730, 2022

    2022

  23. [32]

    Learning optimal power flow value functions with input-convex neural networks

    Andrew Rosemberg, Mathieu Tanneau, Bruno Fanzeres, Joaquim Garcia, and Pascal Van Hen- tenryck. Learning optimal power flow value functions with input-convex neural networks. Electric Power Systems Research, 235:110643, 2024

    2024

  24. [33]

    Sensitivity analysis in nonlinear programs and variational inequalities via continuous selections.SIAM Journal on Control and Optimization, 33(4):1040–1060, 1995

    Jiming Liu. Sensitivity analysis in nonlinear programs and variational inequalities via continuous selections.SIAM Journal on Control and Optimization, 33(4):1040–1060, 1995. 11

    1995

  25. [34]

    Alessandro Soares, Alexandre Street, Tiago Andrade, and Joaquim Dias Garcia. An integrated progressive hedging and benders decomposition with multiple master method to solve the brazilian generation expansion problem.IEEE Transactions on Power Systems, 37(5):4017– 4027, 2022

    2022

  26. [35]

    A three-level static milp model for generation and transmission expansion planning.IEEE Transactions on Power systems, 28(1):202–210, 2012

    David Pozo, Enzo E Sauma, and Javier Contreras. A three-level static milp model for generation and transmission expansion planning.IEEE Transactions on Power systems, 28(1):202–210, 2012

    2012

  27. [36]

    A chance-constrained unit commitment with annk security criterion and significant wind generation.IEEE Transactions on Power systems, 28(3):2842– 2851, 2012

    David Pozo and Javier Contreras. A chance-constrained unit commitment with annk security criterion and significant wind generation.IEEE Transactions on Power systems, 28(3):2842– 2851, 2012

    2012

  28. [37]

    Risk-aware control and optimization for high-renewable power grids.arXiv preprint arXiv:2204.00950, 2022

    Neil Barry, Minas Chatzos, Wenbo Chen, Dahye Han, Chaofan Huang, Roshan Joseph, Michael Klamkin, Seonho Park, Mathieu Tanneau, Pascal Van Hentenryck, et al. Risk-aware control and optimization for high-renewable power grids.arXiv preprint arXiv:2204.00950, 2022

    2022 arXiv

  29. [38]

    Real-time security-constrained economic dispatch and commitment in the pjm: Experiences and challenges

    Simon Tam. Real-time security-constrained economic dispatch and commitment in the pjm: Experiences and challenges. InFERC Software Conference, 2011

    2011

  30. [39]

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

    Wenbo Chen. End-to-end feasible optimization proxies for large-scale economic dispatch. INFORMS, 2023

    2023

  31. [40]

    Luh, Mikhail A

    Xiaorong Sun, Peter B. Luh, Mikhail A. Bragin, Yonghong Chen, Fengyu Wang, and Jie Wan. A decomposition and coordination approach for large-scale security constrained unit commitment problems with combined cycle units. InIEEE Power & Energy Society, pages 1–5, 2017

    2017

  32. [41]

    E. B. Fisher, R. P. O’Neill, and M. C. Ferris. Optimal transmission switching.IEEE Transactions on Power Systems (TPWRS), 23(3):1346–1355, Aug 2008

    2008

  33. [42]

    Transmission expansion planning: A review

    Sumit Verma, Vivekananda Mukherjee, et al. Transmission expansion planning: A review. In3rd International Conference on Energy Efficient Technologies for Sustainability (ICEETS 2016), pages 350–355. IEEE, 2016

    2016

  34. [43]

    Powermodels.jl: An open-source framework for exploring power flow formulations

    Carleton Coffrin, Russell Bent, Kaarthik Sundar, Yeesian Ng, and Miles Lubin. Powermodels.jl: An open-source framework for exploring power flow formulations. In2018 Power Systems Computation Conference (PSCC), pages 1–8, June 2018

    2018

  35. [44]

    Recent iso software enhancements and future software and modeling plans.Federal Energy Regulatory Commission, Tech

    Richard P O’Neill, Thomas Dautel, and Eric Krall. Recent iso software enhancements and future software and modeling plans.Federal Energy Regulatory Commission, Tech. Rep, 2011

    2011

  36. [45]

    Self-supervised primal-dual learning for constrained optimization

    Seonho Park and Pascal Van Hentenryck. Self-supervised primal-dual learning for constrained optimization. InProceedings of the AAAI Conference on Artificial Intelligence, volume 37.4, pages 4052–4060, 2023

    2023

  37. [46]

    High- fidelity machine learning approximations of large-scale optimal power flow.arXiv preprint arXiv:2006.16356, 2020

    Minas Chatzos, Ferdinando Fioretto, Terrence WK Mak, and Pascal Van Hentenryck. High- fidelity machine learning approximations of large-scale optimal power flow.arXiv preprint arXiv:2006.16356, 2020

    2006 arXiv

  38. [47]

    The power grid library for benchmarking ac optimal power flow algorithms.arXiv preprint arXiv:1908.02788, 2019

    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 pr...

    1908 arXiv

  39. [48]

    Dual conic proxies for ac optimal power flow.Electric Power Systems Research, 236:110661, 2024

    Guancheng Qiu, Mathieu Tanneau, and Pascal Van Hentenryck. Dual conic proxies for ac optimal power flow.Electric Power Systems Research, 236:110661, 2024

    2024

  40. [49]

    Learning-based online op- timization for autonomous mobility-on-demand fleet control.arXiv preprint arXiv:2302.03963, 2023

    Kai Jungel, Axel Parmentier, Maximilian Schiffer, and Thibaut Vidal. Learning-based online op- timization for autonomous mobility-on-demand fleet control.arXiv preprint arXiv:2302.03963, 2023

    2023 arXiv

  41. [50]

    Julia: A fresh approach to numerical computing.SIAM Review, 59(1):65–98, 2017

    Jeff Bezanson, Alan Edelman, Stefan Karpinski, and Viral B Shah. Julia: A fresh approach to numerical computing.SIAM Review, 59(1):65–98, 2017

    2017

  42. [51]

    12 A AC-OPF Nomenclature This appendix lists every symbol and constraint that appears in the AC optimal power-flow formulation of Model 1

    PACE.Partnership for an Advanced Computing Environment (PACE), 2017. 12 A AC-OPF Nomenclature This appendix lists every symbol and constraint that appears in the AC optimal power-flow formulation of Model 1. A.1 Sets and Indices Symbol Definition N Set of buses (nodes) in the ...

    2017

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.