Pith. sign in

REVIEW 3 major objections 5 minor 43 references

A Learning-Based Inexact ADMM for Solving Quadratic Programs

T0 review · 3 major / 5 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 A neural-approximated ADMM provably converges to optimal solutions of convex quadratic programs when learned iterates satisfy specified inexactness conditions.

desk verdict Interesting L2O-ADMM idea with a real theory-implementation gap: the convergence theorem's hypotheses fail on the paper's own Random QP benchmark, and the GPU-vs-CPU speedups need a caveat. read the letter →

arxiv 2505.09391 v1 pith:HZXQ4G7A submitted 2025-05-14 math.OC

Xi Gao , Jinxin Xiong , Linxin Yang , Akang Wang , Weiwei Xu , Jiang Xue This is my paper · ORCID
classification math.OC MSC 90C2090C2568T07
keywords quadraticprogramslearningtooptimizeinexactalternatingdirectionmethodofmultiplierslongshort-termmemorynetworksself-supervisedconvergenceguaranteesADMM
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that a learning-augmented alternating direction method of multipliers (ADMM) can replace the exact subproblem solves with approximations produced by a recurrent neural network and still converge to an optimal solution of a convex quadratic program. The authors cast the learned solver as an inexact ADMM, prove an energy-descent convergence theorem under five explicit conditions on the neural iterates, and add a short exact-ADMM refinement stage to restore feasibility. If the central claim is right, learning-based acceleration of ADMM does not have to give up convergence guarantees: approximate subproblems, no per-iteration factorization, and a provable path to optimality can coexist.

What carries the argument

The load-bearing machinery is the inexact ADMM framework for the two-block reformulation min (1/2)x^T Q x + p^T x + I_{l≤z≤u}(z) subject to Ax=z, where the x- and z-subproblems are solved only approximately. The five conditions are sufficient decrease of the augmented Lagrangian, gradient and subgradient bounds proportional to iterate increments, and a line-search decrease on the final update; condition (17) controls the relaxation step. The convergence proof works through an energy function $E^k = L_\rho(x^k,z^k,y^k) + \Gamma^k$, where $\Gamma^k$ is a weighted sum of squared iterate increments, and the matrix inequalities (31)-(33) make the energy decrease monotonically. From that descent, the paper derives $O(1/K)$ decay of the squared residuals and $O(1/\sqrt{K})$ decay of the subdifferential distance, giving sublinear convergence to an optimal solution.

What would settle it

Choose a Random QP instance from the paper's benchmark, run Algorithm 2, and evaluate conditions (14) and (16) after the first 50 iterations; if either inequality is violated at a late iteration where the algorithm still reports convergence, then Theorem 1's hypotheses fail for that run and any observed convergence is not explained by the theorem.

Watch

Extended reading notes

Core claim

The central claim is Theorem 1: under Assumption 1 (Range(A)=R^m) and conditions (13)-(16), with the line-search condition (17), if the energy sequence is bounded below and matrix inequalities (31)-(33) hold, then Algorithm 2 generates iterates whose Lagrangian, augmented Lagrangian, and energy sequences converge to a common limit F*, whose distance to the Lagrangian subdifferential goes to zero, and whose limit points are optimal solutions of the original QP. This is stated as the learned method maintaining primal-dual convergence while satisfying residual thresholds. The theorem is the bridge that turns the neural solver into a provably convergent optimization algorithm, provided the neural outputs obey the inexact-ADMM conditions at every iteration.

Load-bearing premise

The load-bearing premise is that the neural network outputs satisfy conditions (13)-(16) and the line-search condition (17) at every iteration; the paper's own Section 5.3.2 reports significant violations of these conditions on Random QP instances.

Editorial extensions

If this is right

  • If the trained network satisfies the hypotheses at every iteration, the learned solver is guaranteed to drive primal and dual residuals to zero, so it can be used as a convergent approximate QP solver without per-iteration factorization.
  • Because only approximate subproblem solves are required, each iteration avoids the matrix factorization that dominates classical solvers; the reported wall-clock speedups come from the resulting cheaper iterations.
  • The sublinear rates ($O(1/K)$ on residuals, $O(1/\sqrt{K})$ on subdifferential distance) make the guarantee practical for moderate-accuracy regimes rather than for high-precision asymptotics.
  • The two-stage refinement, which runs a short exact-ADMM phase with a single factorization, restores exact feasibility and brings objective values in line with established solvers, so the neural phase only needs to land in a neighborhood of the solution.
  • The stopping criterion based on the composite residual $R^k$ is justified by the theorem: as $R^k$ goes to zero, the iterates approach a stationary point, which is globally optimal by convexity.

Reading between the lines

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

  • Editorial inference: if the sufficient-decrease and gradient conditions were enforced as constraints or penalty terms during training, the deployed network would satisfy the theorem's hypotheses by construction, closing the theory-practice gap the paper reports for Random QP instances.
  • Editorial inference: the proof is largely agnostic to the specific neural architecture, so any learned oracle that meets conditions (13)-(17) inherits the convergence claim; the result can be read as a certification template for other learned ADMM subproblem solvers.
  • Editorial inference: Assumption 1 excludes rank-deficient constraint matrices, so a natural stress test is to apply the method to QPs with redundant constraints, where the dual sequence may drift; constraint preprocessing or dual normalization would be a testable fix.
  • Editorial inference: the theorem's convergence guarantee concerns the idealized iterates satisfying all five conditions, not the trained model as such; users should monitor the violation of (14) and (16) online and switch to the refinement stage when violations persist.
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

3 major / 5 minor

Summary. The paper proposes I-ADMM-LSTM, a self-supervised LSTM-based inexact ADMM for convex quadratic programs. The method replaces the exact ADMM subproblem solves with LSTM-generated approximate solutions, trains the network by minimizing primal and dual residuals over a window of iterations, and optionally adds a feasibility-restoration stage. The theoretical contribution is a convergence theorem (Theorem 1) showing that, if the learned iterates satisfy the sufficient-decrease and subgradient-boundedness conditions (13)-(16) and the line-search condition (17), together with an energy-boundedness assumption and certain matrix inequalities, then the ADMM iterates converge sublinearly to an optimal solution. The experimental section reports large speedups over Gurobi, SCS, and OSQP on several QP benchmarks, with an empirical validation section assessing whether the theoretical conditions hold.

Significance. If the conditional convergence theorem is sound, it would provide a useful template for learning-based inexact ADMM with provable guarantees under verifiable conditions. The LSTM architecture is parameter-efficient, the self-supervised loss avoids labeled data, and the code is publicly available. The empirical results on several benchmarks are encouraging. However, the paper's central advertised claim, that the implemented method maintains provable primal-dual convergence, is not established: the training loss and Algorithm 2 do not enforce the theorem's hypotheses, the paper's own Section 5.3.2 reports significant violations of those hypotheses on the Random QP benchmark, and some of the theory's assumptions are impossible on that same benchmark. The conditional theorem itself appears to be a genuine sufficient-condition result rather than a tautology, but it does not cover the method as implemented.

major comments (3)
  1. [§5.3.2, §3.1, Algorithm 2] The convergence guarantee in Theorem 1 is conditional on Conditions (13)-(16) and the line-search condition (17) holding at every iteration, but the self-supervised loss (23) and Algorithm 2 do not enforce these conditions, and Section 5.3.2 states that 'significant violations occur for Random QP instances' and that there is 'a subtle yet critical gap between inexact computation and provable convergence.' Therefore the abstract's claim that 'near-perfect compliance with the theoretical assumptions, consequently ensuring algorithm convergence' is not supported by the paper's own experiments.
  2. [§4, Assumption 1; Table 3] Assumption 1 requires Range(A)=R^m, i.e., A must have full row rank. For the Random QP benchmark, Table 3 reports n=1000 and m=2000, so A is a 2000×1000 matrix and its range cannot equal R^2000. Thus Theorem 1 cannot cover the Random QP results reported in Table 2, and no verification of full row rank is provided for any of the other datasets.
  3. [§4 vs §3.2, Eqs. (25), (28)-(30)] Proposition 1 and Theorem 1 are proved for a constant scalar penalty parameter ρ appearing in the energy definitions (28)-(30), in the conditions (13)-(16), and in the dual update y^{k+1}=y^k+ρ(A\tilde{x}^{k+1}-z^{k+1}) used in the proof of Lemma 1. Algorithm 2, however, uses an iteration-dependent diagonal penalty matrix ρ^{k+1} (Eq. (25)) and updates y using \tilde{z}^{k+1} reconstructed from the approximate solution of the linear system (19), which need not equal A\tilde{x}^{k+1} unless that linear system is solved exactly. No condition among (13)-(17) controls the mismatch A\tilde{x}^{k+1}-\tilde{z}^{k+1}, and no argument is given for the time-varying ρ; as stated, Theorem 1 therefore does not apply to Algorithm 2.
minor comments (5)
  1. [Appendix C.2, C.3] The proof contains two visible typos: 'by (17) and , yields' in C.2 is missing a condition or reference, and 'The monotonicity and boundedness of {\tilde{E}^k} is a ensure' in C.3 is grammatically corrupted. Please correct these.
  2. [Table 3] The |IL| and |IU| columns are reported as zero for every instance, including problems that clearly have finite lower and upper bounds (e.g., Random QP with l≤Ax≤u and SVM). This is inconsistent with the dataset descriptions in Appendix D and should be clarified.
  3. [Abstract and §5.3.2] The abstract's phrase 'near-perfect compliance with the theoretical assumptions' is contradicted by Section 5.3.2's report of 'significant violations' on Random QP. The abstract should be reworded to state the conditional nature of the guarantee and the observed violations.
  4. [Algorithm 1, line 4] Algorithm 1 uses \tilde{z}^{k+1} in the dual update, but \tilde{z}^{k+1} is never defined in Section 2.3. It should be defined as A\tilde{x}^{k+1} (or explicitly related to Section 3.1's reconstruction) so that the algorithm matches the theoretical dual update used in the proof.
  5. [Figure 5] Section 5.3.2 describes condition-compliance behavior qualitatively (e.g., 'rare violations in early iterations' and 'significant violations' for Random QP) but does not report quantitative counts or aggregate statistics over the 50 test instances. Adding fractions of violating iterations would strengthen the empirical validation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Theorem 1 is a conditional sufficient-condition proof; the learned-output conditions are stated hypotheses, not restatements of the convergence conclusion.

full rationale

The derivation chain is non-circular. Theorem 1 and Proposition 1 prove energy descent (34)-(35) from explicit hypotheses: Assumption 1, Conditions (13)-(16), boundedness below of Etilde, and matrix inequalities (31)-(33). These hypotheses are not equivalent to the conclusion: the proof via Lemma 1 and the energy decomposition in Appendix C is a genuine sufficient-condition argument, and the constants beta_x, beta_z in (39) are chosen to satisfy (31)-(33), which is legitimate parameter selection rather than a fitted prediction. The LSTM is trained with the residual loss (23), none of whose terms encodes the convergence conclusion; no fitted parameter is renamed as a prediction. Self-citations (e.g., Gao et al. 2024) are motivational and are not load-bearing; the convergence proof rests on the paper's own inequalities and on standard background citations (Xie 2018; Bai et al. 2025). The disclosed failure of the hypotheses on Random QP (Section 5.3.2: 'significant violations occur for Random QP instances') and the impossibility of Assumption 1 when m>n are correctness or applicability concerns about whether the theorem covers the reported experiments; they are not circular reductions, because the theorem honestly states those conditions as assumptions.

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

The central claim rests on two kinds of external inputs: the inexact ADMM conditions from Bai et al. (2025) and Xie (2018), which the paper adapts, and the assumption that the trained LSTM stays within those error conditions. The latter is not guaranteed by the learning procedure and is only empirically checked. The problem-class assumption Assumption 1 (full row rank of A) is stated but not validated on the synthetic datasets. The trainable rho, alpha and the hand-picked constants in Section 5.3.2 are additional degrees of freedom that connect the implementation to the theory. No new physical or formal entities are introduced.

free parameters (4)
  • rho_k^i (penalty parameter per constraint, per iteration) = sigma(rho_bar_k) for inactive constraints; 10^3 * sigma(rho_bar_k) for active constraints (rho_bar_k trainable)
    Learned via the self-supervised loss (23) and used in Algorithm 2. The value of rho affects the subproblem conditioning and the sufficient decrease conditions (13) and (15) that connect the LSTM outputs to the convergence theorem.
  • alpha_k (relaxation parameter per iteration) = 2 * sigma(alpha_bar_k) with alpha_bar_k trainable
    Learned via (23). Appears in the line search condition (17) and the extrapolation step of Algorithm 2. The paper presents it as learnable to avoid heuristic line search.
  • beta_x and beta_z (sufficient decrease constants) = beta_x = 2(1+tau)/(1-tau) [2(sigma_max^Q / rho + c_x)^2 + 8 / sigma_min^{A^T A}], beta_z = 32(1+tau) kappa(A^T…
    Chosen by hand in (39) to satisfy the matrix inequalities (31)-(33). They appear in the sufficient decrease conditions (13) and (15) that the learned iterates must satisfy, so they are part of the link between the implementation and the theory.
  • c_x, c_z, delta, tau (theory constants) = c_x=1.0, c_z=1.0, delta=0.9, tau=0.1
    Hand-picked in Section 5.3.2 for the empirical compliance check. The theorem holds for any valid constants, but the reported 'near-perfect compliance' applies only to these values.
assumptions (4)
  • domain assumption For all z in R^m, z in Range(A) (Assumption 1)
    Used in Section 4 to bound the dual residual term via sigma_min^{A^T A} (eq. 27). It requires A to have full row rank, which can fail for QPs with redundant constraints; the experiments do not verify this condition on each dataset.
  • ad hoc to paper The learned iterates satisfy the sufficient decrease and gradient conditions (13)-(16) and the line search condition (17) at every iteration
    This is the bridge between the LSTM solver and the convergence theorem. It is not enforced by the loss (23) and is only checked empirically; Section 5.3.2 reports violations on Random QP, so the theorem does not cover the implemented method on all benchmarks.
  • domain assumption The energy sequence {Etilde^k} is bounded below
    Assumed in Theorem 1 to ensure the energy sequences converge. Plausible for bounded feasible sets, but not established for the learned trajectories on unbounded QP instances.
  • standard math Q is positive semidefinite and nabla f is Lipschitz continuous with constant sigma_max^Q
    Defines the convex QP class; Lipschitz continuity follows from finiteness of sigma_max of the fixed matrix Q.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Learning-Based Inexact ADMM for Solving Quadratic Programs." pith.science (2026). https://pith.science/paper/HZXQ4G7A

@misc{pith2026250509391,
  author       = {Pith},
  title        = {Pith review of: A Learning-Based Inexact ADMM for Solving Quadratic Programs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HZXQ4G7A}},
  note         = {Machine review of arXiv:2505.09391}
}
abstract

Convex quadratic programs (QPs) constitute a fundamental computational primitive across diverse domains including financial optimization, control systems, and machine learning. The alternating direction method of multipliers (ADMM) has emerged as a preferred first-order approach due to its iteration efficiency - exemplified by the state-of-the-art OSQP solver. Machine learning-enhanced optimization algorithms have recently demonstrated significant success in speeding up the solving process. This work introduces a neural-accelerated ADMM variant that replaces exact subproblem solutions with learned approximations through a parameter-efficient Long Short-Term Memory (LSTM) network. We derive convergence guarantees within the inexact ADMM formalism, establishing that our learning-augmented method maintains primal-dual convergence while satisfying residual thresholds. Extensive experimental results demonstrate that our approach achieves superior solution accuracy compared to existing learning-based methods while delivering significant computational speedups of up to $7\times$, $28\times$, and $22\times$ over Gurobi, SCS, and OSQP, respectively. Furthermore, the proposed method outperforms other learning-to-optimize methods in terms of solution quality. Detailed performance analysis confirms near-perfect compliance with the theoretical assumptions, consequently ensuring algorithm convergence.

Figures

Figures reproduced from arXiv: 2505.09391 by the authors.

Figure 1
Figure 1. An illustration of the I-ADMM-LSTM approach. 3.1 I-ADMM-LSTM To efficiently obtain inexact solutions for subproblem (3), we integrate learning-based techniques for solving its optimality conditions (7). To balance computational efficiency and numerical stability, we reformulate (7) into a condensed system:   Q + σI A⊤ A −ρ −1 I     x˜ k+1 ν k+1   =   σxk − p z k − ρ −1y k   , (19) with ˜z k+1 being recon… view at source ↗
Figure 2
Figure 2. Coordinate-wise LSTM architecture for solving (21), where “L” represents the affine transfor￾mation layer. To enhance computational efficiency, our framework incorporates a single coordinate-wise LSTM cell (Online Appendix A) to generate approximate solutions for (21), with its architecture detailed in [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. An illustration of the I-ADMM-LSTM-FR approach. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Convergence characteristics of I-ADMM-LSTM (-FR) across representative instances, illustrating objective value trajectories, linear system residuals, and primal/dual residual dynamics. δ = 0.9 and τ = 0.1, with βx and βz derived via: βx = 2(1 + τ ) 1 − τ    2 σmax …
Figure 5
Figure 5. Figure 5: The satisfaction of conditions (13) – (17) across I-ADMM-LSTM iterations. 24 [PITH_FULL_IMAGE:figures/full_fig_p024_5.png]
Figure 6
Figure 6. Figure 6: The time required for I-ADMM-LSTM to solve multiple optimization problems in parallel. 39 [PITH_FULL_IMAGE:figures/full_fig_p039_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 27 canonical work pages

  1. [1]

    Learning to learn by gradient descent by gradient descent

    Marcin Andrychowicz, Misha Denil, Sergio Gomez, Matthew W Hoffman, David Pfau, Tom Schaul, Brendan Shillingford, and Nando De Freitas. Learning to learn by gradient descent by gradient descent. Advances in neural information processing systems, 29, 2016

    2016

  2. [2]

    Practical large-scale linear programming using primal-dual hybrid gradient

    David Applegate, Mateo D \' az, Oliver Hinder, Haihao Lu, Miles Lubin, Brendan O'Donoghue, and Warren Schudy. Practical large-scale linear programming using primal-dual hybrid gradient. Advances in Neural Information Processing Systems, 34: 0 20243--20257, 2021

    2021

  3. [3]

    Mosek optimization toolbox for matlab

    Mosek ApS. Mosek optimization toolbox for matlab. User’s Guide and Reference Manual, Version, 4 0 (1), 2019

    work page 2019

  4. [4]

    An inexact admm for separable nonconvex and nonsmooth optimization

    Jianchao Bai, Miao Zhang, and Hongchao Zhang. An inexact admm for separable nonconvex and nonsmooth optimization. Computational Optimization and Applications, pages 1--35, 2025

    work page 2025

  5. [5]

    Infeasibility detection in the alternating direction method of multipliers for convex optimization

    Goran Banjac, Paul Goulart, Bartolomeo Stellato, and Stephen Boyd. Infeasibility detection in the alternating direction method of multipliers for convex optimization. Journal of Optimization Theory and Applications, 183: 0 490--519, 2019

    work page 2019

  6. [6]

    Inexact interior-point method

    Stefania Bellavia. Inexact interior-point method. Journal of Optimization Theory and Applications, 96: 0 109--121, 1998

    work page 1998

  7. [7]

    Predictive control for linear and hybrid systems

    Francesco Borrelli, Alberto Bemporad, and Manfred Morari. Predictive control for linear and hybrid systems. Cambridge University Press, 2017

    work page 2017

  8. [8]

    Distributed optimization and statistical learning via the alternating direction method of multipliers

    Stephen Boyd, Neal Parikh, Eric Chu, Borja Peleato, Jonathan Eckstein, et al. Distributed optimization and statistical learning via the alternating direction method of multipliers. Foundations and Trends in Machine learning , 3 0 (1): 0 1--122, 2011

    2011

Show all 43 references
  1. [9]

    Performance bounds and suboptimal policies for multi--period investment

    Stephen Boyd, Mark T Mueller, Brendan O’Donoghue, Yang Wang, et al. Performance bounds and suboptimal policies for multi--period investment. Foundations and Trends in Optimization , 1 0 (1): 0 1--72, 2013

    2013

  2. [10]

    On representing mixed-integer linear programs by graph neural networks

    Ziang Chen, Jialin Liu, Xinshang Wang, and Wotao Yin. On representing mixed-integer linear programs by graph neural networks. In The Eleventh International Conference on Learning Representations, 2022

    2022

  3. [11]

    Support-vector networks

    Corinna Cortes. Support-vector networks. Machine Learning, 1995

    1995

  4. [12]

    Dc3: A learning method for optimization with hard constraints

    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]

    Globally convergent inexact newton methods

    Stanley C Eisenstat and Homer F Walker. Globally convergent inexact newton methods. SIAM Journal on Optimization, 4 0 (2): 0 393--422, 1994

    1994

  6. [14]

    Ipm-lstm: A learning-based interior point method for solving nonlinear programs

    Xi Gao, Jinxin Xiong, Akang Wang, Jiang Xue, Qingjiang Shi, et al. Ipm-lstm: A learning-based interior point method for solving nonlinear programs. Advances in Neural Information Processing Systems, 37: 0 122891--122916, 2024

    2024

  7. [15]

    Model predictive control: Theory and practice—a survey

    Carlos E Garcia, David M Prett, and Manfred Morari. Model predictive control: Theory and practice—a survey. Automatica, 25 0 (3): 0 335--348, 1989

    1989

  8. [16]

    The machine learning for combinatorial optimization competition (ml4co): Results and insights

    Maxime Gasse, Simon Bowly, Quentin Cappart, Jonas Charfreitag, Laurent Charlin, Didier Ch \'e telat, Antonia Chmiela, Justin Dumouchelle, Ambros Gleixner, Aleksandr M Kazachkov, et al. The machine learning for combinatorial optimization competition (ml4co): Results and insight...

    2021

  9. [17]

    Learning fast approximations of sparse coding

    Karol Gregor and Yann LeCun. Learning fast approximations of sparse coding. In Proceedings of the 27th international conference on international conference on machine learning, pages 399--406, 2010

    2010

  10. [18]

    Bounds on eigenvalues of matrices arising from interior-point methods

    Chen Greif, Erin Moulding, and Dominique Orban. Bounds on eigenvalues of matrices arising from interior-point methods. SIAM Journal on Optimization, 24 0 (1): 0 49--83, 2014

    2014

  11. [19]

    Gurobi Optimizer Reference Manual, 2023

    LLC Gurobi Optimization. Gurobi Optimizer Reference Manual, 2023. URL https://www.gurobi.com/documentation/

    2023

  12. [20]

    Alternating direction method with self-adaptive penalty parameters for monotone variational inequalities

    Bing-Sheng He, Hai Yang, and SL Wang. Alternating direction method with self-adaptive penalty parameters for monotone variational inequalities. Journal of Optimization Theory and applications, 106: 0 337--356, 2000

    2000

  13. [21]

    On the convergence of primal-dual hybrid gradient algorithm

    Bingsheng He, Yanfei You, and Xiaoming Yuan. On the convergence of primal-dual hybrid gradient algorithm. SIAM Journal on Imaging Sciences, 7 0 (4): 0 2526--2537, 2014

    2014

  14. [22]

    Robust statistics

    Peter J Huber and Elvezio M Ronchetti. Robust statistics. John Wiley & Sons, 2011

    2011

  15. [23]

    Gonzalez, Ion Stoica, and Ken Goldberg

    Jeffrey Ichnowski, Paras Jain, Bartolomeo Stellato, Goran Banjac, Michael Luo, Francesco Borrelli, Joseph E. Gonzalez, Ion Stoica, and Ken Goldberg. Accelerating quadratic optimization with reinforcement learning, 2021

    2021

  16. [24]

    Implementation of warm-start strategies in interior-point methods for linear programming in fixed dimension

    Elizabeth John and E Alper Y ld r m. Implementation of warm-start strategies in interior-point methods for linear programming in fixed dimension. Computational Optimization and Applications, 41 0 (2): 0 151--183, 2008

    2008

  17. [25]

    Self-supervised equality embedded deep lagrange dual for approximate constrained optimization

    Hongseok Kim et al. Self-supervised equality embedded deep lagrange dual for approximate constrained optimization. arXiv preprint arXiv:2306.06674, 2023

    2023 arXiv

  18. [26]

    Adam: a method for stochastic optimization

    DP Kingma. Adam: a method for stochastic optimization. In Int Conf Learn Represent, 2014

    2014

  19. [27]

    Pdhg-unrolled learning-to-optimize method for large-scale linear programming

    Bingheng Li, Linxin Yang, Yupeng Chen, Senmiao Wang, Haitao Mao, Qian Chen, Yao Ma, Akang Wang, Tian Ding, Jiliang Tang, et al. Pdhg-unrolled learning-to-optimize method for large-scale linear programming. In Proceedings of the 41st International Conference on Machine Learning...

    2024

  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, 2023

    2023

  21. [29]

    Low complexity homeomorphic projection to ensure neural-network solution feasibility for optimization over (non-) convex set

    Enming Liang, Minghua Chen, and Steven Low. Low complexity homeomorphic projection to ensure neural-network solution feasibility for optimization over (non-) convex set. In Conference on Parsimony and Learning (Recent Spotlight Track), 2023

    2023

  22. [30]

    Splitting algorithms for the sum of two nonlinear operators

    Pierre-Louis Lions and Bertrand Mercier. Splitting algorithms for the sum of two nonlinear operators. SIAM Journal on Numerical Analysis, 16 0 (6): 0 964--979, 1979

    1979

  23. [31]

    Towards constituting mathematical structures for learning to optimize

    Jialin Liu, Xiaohan Chen, Zhangyang Wang, Wotao Yin, and HanQin Cai. Towards constituting mathematical structures for learning to optimize. In International Conference on Machine Learning, pages 21426--21449. PMLR, 2023

    2023

  24. [32]

    A practical and optimal first-order method for large-scale convex quadratic programming

    Haihao Lu and Jinwen Yang. A practical and optimal first-order method for large-scale convex quadratic programming. arXiv preprint arXiv:2311.07710, 2023

    2023 arXiv

  25. [33]

    Solving mixed integer programs using neural networks

    Vinod Nair, Sergey Bartunov, Felix Gimeno, Ingrid Von Glehn, Pawel Lichocki, Ivan Lobov, Brendan O'Donoghue, Nicolas Sonnerat, Christian Tjandraatmadja, Pengming Wang, et al. Solving mixed integer programs using neural networks. arXiv preprint arXiv:2012.13349, 2020

    2012 arXiv

  26. [34]

    Operator splitting for a homogeneous embedding of the linear complementarity problem

    Brendan O'Donoghue. Operator splitting for a homogeneous embedding of the linear complementarity problem. SIAM Journal on Optimization, 31 0 (3): 0 1999--2023, 2021

    1999

  27. [35]

    Conic optimization via operator splitting and homogeneous self-dual embedding

    Brendan O’donoghue, Eric Chu, Neal Parikh, and Stephen Boyd. Conic optimization via operator splitting and homogeneous self-dual embedding. Journal of Optimization Theory and Applications, 169: 0 1042--1068, 2016

    2016

  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]

    End-to-end learning to warm-start for real-time quadratic optimization

    Rajiv Sambharya, Georgina Hall, Brandon Amos, and Bartolomeo Stellato. End-to-end learning to warm-start for real-time quadratic optimization. In Learning for Dynamics and Control Conference, pages 220--234. PMLR, 2023

    2023

  30. [38]

    Osqp: An operator splitting solver for quadratic programs

    Bartolomeo Stellato, Goran Banjac, Paul Goulart, Alberto Bemporad, and Stephen Boyd. Osqp: An operator splitting solver for quadratic programs. Mathematical Programming Computation, 12 0 (4): 0 637--672, 2020

    2020

  31. [39]

    Deep admm-net for compressive sensing mri

    Jian Sun, Huibin Li, Zongben Xu, et al. Deep admm-net for compressive sensing mri. Advances in neural information processing systems, 29, 2016

    2016

  32. [40]

    Regression shrinkage and selection via the lasso

    Robert Tibshirani. Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society Series B: Statistical Methodology, 58 0 (1): 0 267--288, 1996

    1996

  33. [41]

    On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming

    Andreas W \"a chter and Lorenz T Biegler. On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming. Mathematical programming, 106: 0 25--57, 2006

    2006

  34. [42]

    On inexact admms with relative error criteria

    Jiaxin Xie. On inexact admms with relative error criteria. Computational Optimization and Applications, 71 0 (3): 0 743--765, 2018

    2018

  35. [43]

    A fast successive qp algorithm for general mean-variance portfolio optimization

    Shengjie Xiu, Xiwen Wang, and Daniel P Palomar. A fast successive qp algorithm for general mean-variance portfolio optimization. IEEE Transactions on Signal Processing, 2023

    2023

Pith tools

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