Pith. sign in

REVIEW 3 major objections 4 minor 60 references

Deep Distributed Optimization for Large-Scale Quadratic Programming

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper claims that unrolling a distributed OSQP-style ADMM solver into a deep network, trained on small problems, produces a QP solver that scales to 50K variables and 150K constraints with 1.5-3.5 times fewer iterations than its…

desk verdict Promising learning-to-optimize architecture with a broken convergence proof: Lemma 1's polar-cone claim is false, so Theorem 1 is unproven. read the letter →

arxiv 2412.12156 v2 pith:5WXXSHON submitted 2024-12-11 math.OC cs.LGcs.MA

Augustinos D. Saravanos , Hunter Kuperman , Alex Oshin , Arshiya Taj Abdul , Vincent Pacelli , Evangelos A. Theodorou This is my paper · ORCID
classification math.OCcs.LGcs.MA MSC 90C2090C2568T07
keywords quadraticprogrammingdistributedoptimizationADMMdeepunfoldinglearningtooptimizePAC-BayesgeneralizationOSQPconsensus
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 tries to establish that a deep-unrolled version of a newly introduced distributed quadratic programming solver can be trained on small problems and then solve much larger network-structured QPs with fewer iterations and far less wall-clock time than a standard centralized solver. The underlying method, DistributedQP, combines OSQP-style operator splitting with a consensus formulation so that each node solves only local subproblems. Unrolling this optimizer for a fixed number of layers and learning the penalty and over-relaxation parameters as feedback policies yields DeepDistributedQP, which the authors test on random QPs, optimal control, network flow, and distributed LASSO. The authors also claim PAC-Bayes bounds on the expected optimality gap for unseen problems, and they prove convergence of DistributedQP to optimality under iteration-varying local parameters.

What carries the argument

The load-bearing object is the unrolled iteration graph of DistributedQP: local nodes maintain primal variables $(x_i, z_i, s_i)$ and dual variables $(\nu_i, \lambda_i, y_i)$, a consensus step averages local $x_i$ into the global variable $w$ using weights $\mu_i$, and each layer maps local residuals to new penalty parameters through small fully-connected feedback networks. The local update solves a KKT system that is reduced to a positive-definite linear system and solved by conjugate gradients, with gradients through the CG loop obtained by implicit differentiation rather than by unrolling inner iterations. A separate PAC-Bayes argument trains a Gaussian distribution over the policy weights and bounds the expected progress metric using the inverse-KL form in Theorem 3.

What would settle it

Run a one-step numerical check of Lemma 1's relation (R4) on a small inequality-constrained QP: after the projection update (6), compute $\lambda_i^{k+1}$ and evaluate $\lambda_i^{k+1\top}(t_1-t_2)$ for two feasible slack vectors $t_1,t_2$ with $t_j \le b_i$; any nonzero value contradicts (R4) and would falsify the proof of Theorem 1 as written.

Watch

Extended reading notes

Core claim

The central claim is that DeepDistributedQP, trained only on low-dimensional instances, generalizes to problems with up to 50K variables and 150K constraints using the same learned policy, reaching a target accuracy in 1.5-3.5 times fewer iterations than DistributedQP and orders-of-magnitude less wall-clock time than OSQP. The paper further claims that DistributedQP is guaranteed to converge to the optimal solution even when each node uses its own iteration-varying penalty parameters, a property needed for deep unfolding. On top of this, Theorem 3 provides a PAC-Bayes bound on the expected relative progress of the final iterate for unseen problems drawn from the same distribution as the training data, and the experiments report that these bounds are tight relative to actual performance.

Load-bearing premise

The convergence proof of Theorem 1 relies on Lemma 1's relation (R4), which asserts that the updated dual variable $\lambda_i$ is orthogonal to the difference of any two feasible slack vectors; if that orthogonality does not actually follow from the polar-cone argument, the descent inequality (68) and the convergence theorem's proof as written collapse.

Editorial extensions

If this is right

  • Practitioners could train a distributed QP policy once on small problems and deploy it on much larger network-structured QPs without retraining at scale.
  • Closed-loop policies that adapt penalty parameters from local residuals reach target accuracy faster than open-loop or fixed parameters, so learned residual feedback can replace manual ADMM tuning.
  • The PAC-Bayes guarantee, if valid, makes the average final-iterate progress of the learned solver certifiable with high probability for unseen problems from the same distribution.
  • Because the underlying method is a two-block ADMM scheme, the same unrolling recipe is a direct candidate for other linearly constrained convex problems beyond quadratic programming.

Reading between the lines

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

  • The same unrolling-and-feedback recipe could be transplanted to other two-block ADMM solvers with projection steps, though the paper does not claim this extension itself.
  • The reported wall-clock advantage over OSQP depends on the conjugate-gradient indirect solve and GPU parallelism; on very small problems or CPU-only hardware, the comparison curve could shift, which is a testable extension of the paper's scaling claims.
  • The PAC-Bayes guarantee is distribution-dependent, so applying the trained policy to problem classes far from the training distribution falls outside the certified regime; a natural next experiment would measure how the bound degrades under distribution shift.
  • Because training uses the known optimal solution as supervision, a self-supervised variant that trains on residual norms instead of optimality gaps would remove the need for pre-solved datasets, which the paper notes only as future work.
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 proposes DistributedQP, a distributed ADMM/OSQP hybrid for convex QPs with a consensus structure, and proves a convergence guarantee for iteration-varying local penalties and over-relaxation (Theorem 1, Appendix C). It then unrolls DistributedQP into a deep network, DeepDistributedQP, with learned feed-forward and feedback policies for the penalty parameters, and a centralized variant DeepQP. The authors derive PAC-Bayes bounds on a relative progress metric for the learned policy and report extensive experiments showing that DeepDistributedQP, trained on small problems, generalizes to much larger problems and requires fewer iterations than DistributedQP and far less wall-clock time than OSQP.

Significance. The paper addresses an important and timely problem: making distributed QP solvers adaptive and scalable through deep unfolding. The empirical study is broad (six problem classes, up to ̃50K variables), and the use of implicit differentiation to backpropagate through the inner linear solver, as well as the PAC-Bayes formulation with a bounded relative-progress loss, are thoughtful components. If the advertised convergence guarantee were proven, this would be a solid contribution to learned distributed optimization. As submitted, however, the central proof of Theorem 1 rests on a false polar-cone identity, so the theoretical foundation of the paper is not established. The empirical results are interesting but do not by themselves compensate for the invalid proof of a headline claim.

major comments (3)
  1. [Appendix C.2, Lemma 1 (R4) and Definition 1] Definition 1 incorrectly defines polar cones by d^T dbar = 0; the polar cone is defined by d^T dbar ≤ 0 for all d in the cone. Consequently, relation (R4), i.e., λ_i^{k+1T}(t1 - t2) = 0 for all t1, t2 ∈ C_i, is false. For C_i = {s | s ≤ b_i}, the shifted cone is Cbar_i = {p | p ≤ 0} and its polar is Cbar_i^o = {p | p ≥ 0}; with b_i = 0, λ_i^{k+1} = 1, t1 = 0 and t2 = -1, the left side of (R4) equals 1, not 0. This is not a cosmetic issue: Lemma 3 uses (R4) at equation (67) to cancel the final term of (R8), and Lemma 4 uses (R4) to assert the signs in (74)-(76) that are needed to derive the descent inequality (68). The summability (88) and the convergence conclusions of Theorem 1 consequently do not follow from the submitted proof.
  2. [Appendix C.2, Lemma 4, inequalities (74)-(76)] Even if one replaces (R4) with the correct polar-cone inequality, the sign claims (74) and (75) are not valid. With C_i = {s ≤ b_i}, KKT condition (44c) yields λ*_i^T(s_i - s*_i) ≤ 0, and λ_i^{k+1} ∈ Cbar_i^o does not imply λ_i^{k+1T}(s_i - s*_i) ≥ 0; a scalar example with b_i = 0, s*_i = -1 (so λ*_i = 0), s_i = -2 and λ_i^{k+1} = 1 gives (λ_i^{k+1} - λ*_i)^T(s_i - s*_i) = -1 < 0. Thus inequality (76) is not established, and the derivation of (78) and (68) in Lemma 4 is invalid.
  3. [Appendix C, proof of Theorem 1] The final step of the proof of Theorem 1 states that relations (R1) and (R4) verify KKT condition (44c)-(44d). Since (R4) is false, the KKT verification of the limit point is also unsupported. The convergence guarantee of DistributedQP is therefore unproven in the submitted manuscript. The authors would need to supply a substantially different proof of convergence, or remove the guarantee and reframe the contribution accordingly.
minor comments (4)
  1. [Equation (16)] The sample-convergence bound in (16) is misstated: D_KL(\hat q_S(P;M) \| M^{-1} log(2/ε)) is not a valid KL divergence because the second argument mixes a probability with a log term; please provide the correct binomial-tail or Bennett-style bound with proper definitions of all quantities.
  2. [Table 6 caption and Section 6.2] For OSQP with the direct method, the reported time is only for the first iteration under the assumption that the KKT factorization is reused for all subsequent iterations; this assumption should be stated in the main text at the point of comparison, as it materially affects the wall-clock claim.
  3. [Section 7, Conclusion] The sentence 'both DeepDistributedQP and Distributed significantly outperform OSQP' is missing 'QP' after 'Distributed'.
  4. [Section 6, first paragraph] The text 'performed on an system' should read 'performed on a system'.

Circularity Check

1 steps flagged · score 6.0 of 10

Theorem 1's convergence proof obtains the key orthogonality (R4) by redefining polar cones to have zero inner products, so the descent inequality and convergence claim reduce to a definitional assumption.

  1. self definitional [Appendix C.2, Definition 1 and Lemma 1, relation (R4) (Eq. (43))]
    "Definition 1 (Polar cones). Two cone sets D and Do are called polar cones if for any d ∈ D and ¯d ∈ Do, if follows that d⊤ ¯d = 0. ... Since ¯Ci and ¯Co i are polar cones, and using (42), the following relation holds true by the definition of polar cones, λk+1 i ⊤(t − bi) = 0 for all t ∈ Ci."

    The paper needs the equality λ_i^{k+1⊤}(t1-t2)=0 in (R4) to make its later descent argument work. Instead of deriving this from the geometry of C_i={s|s≤b_i}, it builds the equality into Definition 1 by asserting that polar cones satisfy d^T d̄=0. The actual polar cone of {p|p≤0} is {p|p≥0}, for which d^T d̄ can be strictly positive; the correct relation is ≤0, not =0. The proof then invokes 'by the definition of polar cones' to conclude (R4), so the desired orthogonality is assumed by construction rather than proven. This is load-bearing: Lemma 3 uses (R4) to eliminate the last term in (66) and obtain (R8); Lemma 4 uses (R4) to assert (74)-(76) and derive the descent inequality (68); Theorem 1 relies on (68) for the summability (88) and for verifying the KKT condition (44c).

full rationale

The empirical learning claims are not internally circular: DeepDistributedQP is trained on small problems with optimality labels produced by OSQP/DistributedQP and evaluated on held-out problems of larger scale, which is external supervision rather than the paper's own output being used as ground truth. The PAC-Bayes argument is standard and uses a separate prior dataset. The acknowledged limitation in Section 6.3—that training requires a dataset of pre-solved problems—is an honest statement of supervised dependence, not a circularity. However, the theoretical convergence proof of DistributedQP contains a definitional circularity: the key relation (R4) is made true by redefining polar cones with equality instead of the correct inequality. Because Lemma 3, Lemma 4, and Theorem 1 all depend on (R4), the claimed convergence guarantee is partially forced by definition. The false polar-cone statement is also a mathematical correctness issue, but the circular aspect is that the proof obtains its needed equality from its own definition rather than from the mathematics of the constraint set.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical or mathematical entities. The central claim rests on standard convex QP assumptions, the consensus mapping G, PAC-Bayes machinery, and the learned policy parameters. The ledger lists the false polar-cone step as an ad hoc assumption because the convergence proof relies on it to derive the descent inequality.

free parameters (3)
  • Per-layer feed-forward penalty parameters bar_rho_i^k, bar_mu_i^k, bar_alpha^k = not reported (learned)
    These are learned by training on a dataset of pre-solved QPs and are the core mechanism by which DeepDistributedQP and DeepQP accelerate convergence.
  • Feedback policy weights theta_i^k = not reported (learned)
    Fully-connected MLP weights in the closed-loop policies; the paper specifies 2x16 hidden units and Adam training with learning rate 1e-3.
  • Training loss schedule gamma_k = exp((k-K)/5)
    Chosen by hand, not fitted; it scales layer contributions in the supervised loss and affects the final learned policy.
assumptions (5)
  • domain assumption Convexity and positive definiteness of Q_i for each local problem
    Problem (2) assumes Q_i is positive definite, which ensures unique local updates and solvability of the KKT systems in Eqs. (4) and (12).
  • domain assumption Full column rank of the consensus mapping G
    Used in Appendix B and Theorem 1 to infer w^{k+1}-w^k goes to zero from G(w^{k+1}-w^k) going to zero; requires every global variable to appear in at least one local variable.
  • ad hoc to paper Polar-cone orthogonality: for polar cones C and C^o, any d in C and dbar in C^o satisfy d^T dbar = 0
    This is the premise behind relation (R4) in Lemma 1. It is not true for general cones; polar cones only guarantee d^T dbar <= 0. The proof of Theorem 1 depends on this step, so it is a load-bearing ad hoc assumption.
  • standard math PAC-Bayes and sample convergence bounds
    Theorem 3 uses standard PAC-Bayes bounds and a sample convergence bound from the cited literature in Appendix G.
  • standard math Implicit Function Theorem
    Theorem 2 uses the standard implicit function theorem to derive gradients through the CG solve in Eq. (12).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deep Distributed Optimization for Large-Scale Quadratic Programming." pith.science (2026). https://pith.science/paper/5WXXSHON

@misc{pith2026241212156,
  author       = {Pith},
  title        = {Pith review of: Deep Distributed Optimization for Large-Scale Quadratic Programming},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5WXXSHON}},
  note         = {Machine review of arXiv:2412.12156}
}
read the original abstract

Quadratic programming (QP) forms a crucial foundation in optimization, encompassing a broad spectrum of domains and serving as the basis for more advanced algorithms. Consequently, as the scale and complexity of modern applications continue to grow, the development of efficient and reliable QP algorithms is becoming increasingly vital. In this context, this paper introduces a novel deep learning-aided distributed optimization architecture designed for tackling large-scale QP problems. First, we combine the state-of-the-art Operator Splitting QP (OSQP) method with a consensus approach to derive DistributedQP, a new method tailored for network-structured problems, with convergence guarantees to optimality. Subsequently, we unfold this optimizer into a deep learning framework, leading to DeepDistributedQP, which leverages learned policies to accelerate reaching to desired accuracy within a restricted amount of iterations. Our approach is also theoretically grounded through Probably Approximately Correct (PAC)-Bayes theory, providing generalization bounds on the expected optimality gap for unseen problems. The proposed framework, as well as its centralized version DeepQP, significantly outperform their standard optimization counterparts on a variety of tasks such as randomly generated problems, optimal control, linear regression, transportation networks and others. Notably, DeepDistributedQP demonstrates strong generalization by training on small problems and scaling to solve much larger ones (up to 50K variables and 150K constraints) using the same policy. Moreover, it achieves orders-of-magnitude improvements in wall-clock time compared to OSQP. The certifiable performance guarantees of our approach are also demonstrated, ensuring higher-quality solutions over traditional optimizers.

Figures

Figures reproduced from arXiv: 2412.12156 by the authors.

Figure 1
Figure 1. Wall-clock time compari￾son. DeepDistributedQP, DistributedQP (ours) and OSQP on large-scale QPs. As the scale of modern decision-making ap￾plications rapidly increases, there is an emerging interest in developing effective optimization ar￾chitectures for addressing high-dimensional prob￾lems. Given the fundamental role of QP in opti￾mization, there is a clear demand for algorithms capable of solving large-scale QPs… view at source ↗
Figure 2
Figure 2. Example of consensus mapping G in problem (2). A convex QP problem is expressed in a general cen￾tralized form as min 1 2 x ⊤Qx + q ⊤x s.t. Ax ≤ b, (1) where x ∈ R n is the decision vector and ζ = {Q ∈ S n ++, q ∈ R n , A ∈ R m×n , b ∈ R m} are the problem data. 1 As the scale of such problems increases to higher dimensions, there is often an underlying net￾worked/decentralized structure that could be lever￾aged for… view at source ↗
Figure 3
Figure 3. The DeepDistributedQP architecture. The proposed framework relies on unrolling the DistributedQP optimizer as a supervised deep learning framework. In partic￾ular, we interpret its iterations (4)-(9) as sequential network layers and introduce learnable components (orange blocks) to facilitate reaching the desired accuracy after a predefined number of allowed iterations. 4.1 Main Architecture Architecture overview. T… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: DeepQP: The centralized version of DeepDistributedQP which boils down to unfolding the standard OSQP method. Training loss. Let S = {ζ j} H j=1 be a dataset consisting of H problem instances ζ j = {(Qi , qi , Ai , bi) N i=1, w∗}j subject to the known mapping G of probl…
Figure 5
Figure 5. Figure 5: Small-scale centralized comparison between DeepQP and OSQP. Across all tested problems, DeepQP consistently outperforms OSQP (same per-iteration complexity using the indirect method). 6.1 Small-Scale Centralized Experiments: DeepQP vs OSQP Setup. We begin with comparin…
Figure 6
Figure 6. Figure 6: Scaling DeepDistributedQP to high-dimensional problems. Left: Com￾parison between DeepDistributedQP and its traditional optimization counterpart Distribut￾edQP (same per-iteration complexity). Right: Total wall-clock time required by DeepDis￾tributedQP, DistributedQP a…
Figure 7
Figure 7. Figure 7: Left: Local vs shared policies. We showcase the advantage of learning local policies over shared ones. Right: Performance guarantees. The obtained generalization bounds guarantee the performance of DeepDistributedQP and its improvements over its standard optimization c…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 52 canonical work pages

  1. [1]

    Agrawal, B

    A. Agrawal, B. Amos, S. Barratt, S. Boyd, S. Diamond, and J. Z. Kolter. Differentiable convex optimization layers. Advances in neural information processing systems, 32, 2019

    work page 2019

  2. [2]

    P. Alquier. User-friendly introduction to PAC - B ayes bounds. Foundations and Trends in Machine Learning, 17 0 (2): 0 174--303, 2024

    work page 2024

  3. [3]

    Amos and J

    B. Amos and J. Z. Kolter. Optnet: Differentiable optimization as a layer in neural networks. In International conference on machine learning, pages 136--145. PMLR, 2017

    work page 2017

  4. [4]

    Amos et al

    B. Amos et al. Tutorial on amortized optimization. Foundations and Trends in Machine Learning , 16 0 (5): 0 592--732, 2023

    work page 2023

  5. [5]

    Belotti, C

    P. Belotti, C. Kirches, S. Leyffer, J. Linderoth, J. Luedtke, and A. Mahajan. Mixed-integer nonlinear optimization. Acta Numerica, 22: 0 1--131, 2013

    2013

  6. [6]

    Biagioni, P

    D. Biagioni, P. Graf, X. Zhang, A. S. Zamzam, K. Baker, and J. King. Learning-accelerated admm for distributed dc optimal power flow. IEEE Control Systems Letters, 6: 0 1--6, 2020

    work page 2020

  7. [7]

    S. Boyd, N. Parikh, E. Chu, B. Peleato, J. 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

    work page 2011

  8. [8]

    S. W. Chen, T. Wang, N. Atanasov, V. Kumar, and M. Morari. Large scale model predictive control with neural networks and primal active sets. Automatica, 135: 0 109947, 2022 a

    work page 2022

Show all 60 references
  1. [9]

    T. Chen, X. Chen, W. Chen, H. Heaton, J. Liu, Z. Wang, and W. Yin. Learning to optimize: A primer and a benchmark. Journal of Machine Learning Research, 23 0 (189): 0 1--59, 2022 b

    2022

  2. [10]

    A. R. Conn, N. I. Gould, and P. L. Toint. Trust region methods. SIAM, 2000

    2000

  3. [11]

    Conte, T

    C. Conte, T. Summers, M. N. Zeilinger, M. Morari, and C. N. Jones. Computational aspects of distributed optimization in model predictive control. In 2012 IEEE 51st IEEE conference on decision and control (CDC), pages 6819--6824. IEEE, 2012 a

    2012

  4. [12]

    Conte, N

    C. Conte, N. R. Voellmy, M. N. Zeilinger, M. Morari, and C. N. Jones. Distributed synthesis and control of constrained linear systems. In 2012 American control conference (ACC), pages 6017--6022. IEEE, 2012 b

    2012

  5. [13]

    u t \"u nc \

    G. Cornuejols, J. Pe \ n a, and R. T \"u t \"u nc \"u . Optimization methods in finance. Cambridge University Press, 2018

    2018

  6. [14]

    Cortes and V

    C. Cortes and V. Vapnik. Support-vector networks. Machine Learning, 1995

    1995

  7. [15]

    Deng and W

    W. Deng and W. Yin. On the global and linear convergence of the generalized alternating direction method of multipliers. Journal of Scientific Computing, 66: 0 889--916, 2016

    2016

  8. [16]

    G. K. Dziugaite and D. M. Roy. Computing nonvacuous generalization bounds for deep (stochastic) neural networks with many more parameters than training data. arXiv preprint arXiv:1703.11008, 2017

    2017 arXiv

  9. [17]

    G. K. Dziugaite, K. Hsu, W. Gharbieh, G. Arpino, and D. Roy. On the role of data in PAC - B ayes bounds. In International Conference on Artificial Intelligence and Statistics, pages 604--612. PMLR, 2021

    2021

  10. [18]

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

    1989

  11. [19]

    Hadou, N

    S. Hadou, N. NaderiAlizadeh, and A. Ribeiro. Stochastic unrolled federated learning. arXiv preprint arXiv:2305.15371, 2023

    2023 arXiv

  12. [20]

    Y. He, Q. Shang, X. Huang, J. Liu, and K. Yuan. A mathematics-inspired learning-to-optimize framework for decentralized optimization. arXiv preprint arXiv:2410.01700, 2024

    2024 arXiv

  13. [21]

    Houska, J

    B. Houska, J. Frasch, and M. Diehl. An augmented lagrangian based algorithm for distributed nonconvex optimization. SIAM Journal on Optimization, 26 0 (2): 0 1101--1127, 2016

    2016

  14. [22]

    Huang, Q

    S. Huang, Q. Wu, S. S. Oren, R. Li, and Z. Liu. Distribution locational marginal pricing through quadratic programming for congestion management in distribution networks. IEEE Transactions on Power Systems, 30 0 (4): 0 2170--2178, 2014

    2014

  15. [23]

    Ichnowski, P

    J. Ichnowski, P. Jain, B. Stellato, G. Banjac, M. Luo, F. Borrelli, J. E. Gonzalez, I. Stoica, and K. Goldberg. Accelerating quadratic optimization with reinforcement learning. Advances in Neural Information Processing Systems, 34: 0 21043--21055, 2021

    2021

  16. [24]

    Kishida, M

    M. Kishida, M. Ogura, Y. Yoshida, and T. Wadayama. Deep learning-based average consensus. IEEE Access, 8: 0 142404--142412, 2020

    2020

  17. [25]

    S. G. Krantz and H. R. Parks. The implicit function theorem: history, theory, and applications. Springer Science & Business Media, 2002

    2002

  18. [26]

    Langford and R

    J. Langford and R. Caruana. (not) bounding the true error. Advances in Neural Information Processing Systems, 14, 2001

    2001

  19. [27]

    Lin, S.-C

    S.-S. Lin, S.-C. Horng, et al. Distributed quadratic programming problems of power systems with continuous and discrete variables. IEEE Transactions on Power Systems, 28 0 (1): 0 472--481, 2012

    2012

  20. [28]

    Majumdar, A

    A. Majumdar, A. Farid, and A. Sonar. PAC - B ayes control: learning policies that provably generalize to novel environments. The International Journal of Robotics Research, 40 0 (2-3): 0 574--593, 2021

    2021

  21. [29]

    Mateos, J

    G. Mateos, J. A. Bazerque, and G. B. Giannakis. Distributed sparse linear regression. IEEE Transactions on Signal Processing, 58 0 (10): 0 5262--5276, 2010

    2010

  22. [30]

    Mattingley and S

    J. Mattingley and S. Boyd. Real-time convex optimization in signal processing. IEEE Signal processing magazine, 27 0 (3): 0 50--61, 2010

    2010

  23. [31]

    Monga, Y

    V. Monga, Y. Li, and Y. C. Eldar. Algorithm unrolling: Interpretable, efficient deep learning for signal and image processing. IEEE Signal Processing Magazine, 38 0 (2): 0 18--44, 2021

    2021

  24. [32]

    J. F. Mota. Communication-efficient algorithms for distributed optimization. PhD thesis, Carnegie Mellon University, 2013

    2013

  25. [33]

    J. F. Mota, J. M. Xavier, P. M. Aguiar, and M. P \"u schel. D-admm: A communication-efficient distributed algorithm for separable optimization. IEEE Transactions on Signal processing, 61 0 (10): 0 2718--2723, 2013

    2013

  26. [34]

    J. F. Mota, J. M. Xavier, P. M. Aguiar, and M. P \"u schel. Distributed optimization with local domains: Applications in mpc and network flows. IEEE Transactions on Automatic Control, 60 0 (7): 0 2004--2009, 2014

    2004

  27. [35]

    Navia-Vazquez, D

    A. Navia-Vazquez, D. Gutierrez-Gonzalez, E. Parrado-Hern \'a ndez, and J. Navarro-Abellan. Distributed support vector machines. IEEE Transactions on Neural Networks, 17 0 (4): 0 1091--1097, 2006

    2006

  28. [36]

    Nesterov and A

    Y. Nesterov and A. Nemirovskii. Interior-point polynomial algorithms in convex programming. SIAM, 1994

    1994

  29. [37]

    Noah and N

    Y. Noah and N. Shlezinger. Distributed learn-to-optimize: Limited communications optimization over networks via deep unfolded distributed admm. IEEE Transactions on Mobile Computing, 2024

    2024

  30. [38]

    Nocedal and S

    J. Nocedal and S. J. Wright. Numerical optimization. Springer, 1999

    1999

  31. [39]

    Ogawa and K

    S. Ogawa and K. Ishii. Deep-learning aided consensus problem considering network centrality. In 2021 IEEE 94th Vehicular Technology Conference (VTC2021-Fall), pages 1--5. IEEE, 2021

    2021

  32. [40]

    Pereira, A

    M. Pereira, A. Saravanos, O. So, and E. Theodorou. Decentralized Safe Multi-agent Stochastic Optimal Control using Deep FBSDEs and ADMM . In Proceedings of Robotics: Science and Systems, New York City, NY, USA, June 2022. doi:10.15607/RSS.2022.XVIII.055

    2022 doi

  33. [41]

    J. B. Rawlings, D. Q. Mayne, M. Diehl, et al. Model predictive control: theory, computation, and design, volume 2. Nob Hill Publishing Madison, WI, 2017

    2017

  34. [42]

    Sambharya and B

    R. Sambharya and B. Stellato. Data-driven performance guarantees for classical and learned optimizers. arXiv preprint arXiv:2404.13831, 2024 a

    2024

  35. [43]

    Sambharya and B

    R. Sambharya and B. Stellato. Learning algorithm hyperparameters for fast parametric convex optimization. arXiv preprint arXiv:2411.15717, 2024 b

    2024 arXiv

  36. [44]

    Sambharya, G

    R. Sambharya, G. Hall, B. Amos, and B. 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

  37. [45]

    Sambharya, G

    R. Sambharya, G. Hall, B. Amos, and B. Stellato. Learning to warm-start fixed-point optimization algorithms. Journal of Machine Learning Research, 25 0 (166): 0 1--46, 2024

    2024

  38. [46]

    A. D. Saravanos, Y. Aoyama, H. Zhu, and E. A. Theodorou. Distributed differential dynamic programming architectures for large-scale multiagent control. IEEE Transactions on Robotics, 2023

    2023

  39. [47]

    Shlezinger, Y

    N. Shlezinger, Y. C. Eldar, and S. P. Boyd. Model-based deep learning: On the intersection of deep learning and optimization. IEEE Access, 10: 0 115384--115398, 2022

    2022

  40. [48]

    Shorinwa, T

    O. Shorinwa, T. Halsted, J. Yu, and M. Schwager. Distributed optimization methods for multi-robot systems: Part 1—a tutorial. IEEE Robotics & Automation Magazine, 2024

    2024

  41. [49]

    Soheili and A

    M. Soheili and A. M. Eftekhari-Moghadam. Dqpfs: Distributed quadratic programming based feature selection for big data. Journal of Parallel and Distributed Computing, 138: 0 1--14, 2020

    2020

  42. [50]

    V. Soltan. Moreau-type characterizations of polar cones. Linear Algebra and its Applications, 567: 0 45--62, 2019. ISSN 0024-3795. doi:https://doi.org/10.1016/j.laa.2019.01.006. URL https://www.sciencedirect.com/science/article/pii/S0024379519300199

    2019 doi

  43. [51]

    Stellato, G

    B. Stellato, G. Banjac, P. Goulart, A. Bemporad, and S. Boyd. OSQP : An operator splitting solver for quadratic programs. Mathematical Programming Computation, 12 0 (4): 0 637--672, 2020

    2020

  44. [52]

    Sucker and P

    M. Sucker and P. Ochs. Pac-bayesian learning of optimization algorithms. In International Conference on Artificial Intelligence and Statistics, pages 8145--8164. PMLR, 2023

    2023

  45. [53]

    Sucker, J

    M. Sucker, J. Fadili, and P. Ochs. Learning-to-optimize with pac-bayesian guarantees: Theoretical considerations and practical implementation. arXiv preprint arXiv:2404.03290, 2024

    2024 arXiv

  46. [54]

    Tibshirani

    R. 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

  47. [55]

    Van Parys and G

    R. Van Parys and G. Pipeleers. Distributed mpc for multi-vehicle systems moving in formation. Robotics and Autonomous Systems, 97: 0 144--152, 2017

    2017

  48. [56]

    H. Wang, Y. Shen, Z. Wang, D. Li, J. Zhang, K. B. Letaief, and J. Lu. Decentralized statistical inference with unrolled graph neural networks. In 2021 60th IEEE Conference on Decision and Control (CDC), pages 2634--2640. IEEE, 2021

    2021

  49. [57]

    P. Wolfe. The simplex method for quadratic programming. Econometrica: Journal of the Econometric Society, pages 382--398, 1959

    1959

  50. [58]

    Z. Xu, M. Figueiredo, and T. Goldstein. Adaptive admm with spectral penalty parameter selection. In Artificial Intelligence and Statistics, pages 718--727. PMLR, 2017

    2017

  51. [59]

    S. Zeng, A. Kody, Y. Kim, K. Kim, and D. K. Molzahn. A reinforcement learning approach to parameter selection for distributed optimal power flow. Electric Power Systems Research, 212: 0 108546, 2022

    2022

  52. [60]

    D. Zhu, T. Xu, and J. Lu. A deep reinforcement learning approach to efficient distributed optimization. arXiv preprint arXiv:2311.08827, 2023

    2023 arXiv

Pith tools

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