Pith. sign in

REVIEW 4 major objections 5 minor 44 references

BPQP: A Differentiable Convex Optimization Framework for Efficient End-to-End Learning

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims that the backward pass through a differentiable convex optimization layer can be rewritten as an equality-constrained quadratic program, making gradient computation a standard QP solve.

desk verdict A genuinely useful reformulation of the backward pass as an equality-constrained QP, with a correct-but-underspecified Theorem 1 and a strong but slightly messy experimental section. read the letter →

arxiv 2411.19285 v2 pith:HRNCYU3O submitted 2024-11-28 cs.LG cs.AIq-fin.PM

Jianming Pan , Zeqi Ye , Xiao Yang , Xu Yang , Weiqing Liu , Lewen Wang , Jiang Bian This is my paper · ORCID
classification cs.LGcs.AIq-fin.PM
keywords differentiableoptimizationlayersconveximplicitdifferentiationquadraticprogrammingKKTconditionsADMMbackpropagationportfolio
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 claims that the backward pass through a differentiable convex optimization layer — normally a costly linear solve with the KKT matrix — can be rewritten as an equality-constrained quadratic program whose optimal solution is exactly the gradient vector needed for backpropagation (Theorem 1). The reformulation decouples the forward and backward passes, so any general-purpose QP solver, in practice a first-order operator-splitting method, can compute gradients. If correct, this makes end-to-end training with optimization layers markedly cheaper: the paper reports total runtime speedups of up to 13.54x, 21.02x, and 1.67x on QP, LP, and SOCP layers, and it demonstrates a 500-asset portfolio optimization task that other differentiable layers could not handle at scale. The sympathetic reading is that gradient computation through KKT conditions is itself a convex QP, and exploiting that structure is enough to turn differentiable optimization from a bottleneck into a practical layer.

What carries the argument

The central object is the KKT matrix of the forward optimization problem, and the key move is to replace its complementarity block — the diagonal matrices $D(\lambda^\star)G(z^\star)$ and $D(g(z^\star))$ — with equality constraints on the active set, the set of inequality constraints that hold with equality at the optimum. Because the active set is known after the forward pass, the backward gradient vector solves the equality-constrained quadratic program in Eq. (6), whose KKT conditions reproduce the desired linear system exactly. This hands the backward pass to any QP solver; the paper uses a first-order ADMM solver with sparsity, solution polishing, and active-set machinery, plus a small regularization $\delta I$ to handle redundant constraints.

What would settle it

Construct a small QP whose optimum has an inequality constraint with a dual multiplier comparable to the solver's tolerance (for example $10^{-6}$), then compute the backward gradient both by the paper's Eq. (6) and by a high-precision solution of Eq. (3). If the two disagree beyond numerical noise, or if changing the solver tolerance flips the active set and produces a discontinuous gradient, the exact-equivalence premise of Theorem 1 fails in a practically reachable regime.

Watch

Extended reading notes

Core claim

The central claim is Theorem 1. Once the forward pass has identified the active set of inequality constraints, the vector $[\tilde z, \tilde\lambda, \tilde\nu]$ that solves the backward KKT system (Eq. 3) is exactly the optimal solution of the equality-constrained quadratic program (Eq. 6) with objective $\tfrac12 \tilde z^\top P' \tilde z + q'^\top \tilde z$ and constraints $A'\tilde z = b'$, $G'_+\tilde z = c'_+$. The complementarity block of the KKT matrix is replaced by the active-set equality rows, which turns the backward system into a plain convex QP that any QP solver can handle. To guard against a singular KKT matrix, the implementation adds a small regularization to the diagonal. On simulated problems the resulting gradients match a high-precision KKT solve with 0.992 cosine similarity for QP, while total runtime is reduced by up to 21x on LP layers.

Load-bearing premise

The equivalence in Theorem 1 rests on knowing the exact active set and on every active inequality having a strictly positive dual multiplier; a first-order solver returns only a finite-tolerance solution, so on weakly active or degenerate constraints the active set is approximate and the backward QP can return a different gradient from the exact KKT system.

Editorial extensions

If this is right

  • End-to-end training of networks with optimization layers becomes practical at problem scales where direct KKT inversion is prohibitively slow, such as the reported 500-asset portfolio optimization task.
  • Because the backward pass is a standard convex QP, future improvements in QP solvers translate directly into faster differentiable layers without changing the method.
  • The forward solver no longer needs to be differentiable or share factorizations with the backward pass, so each pass can use the best available algorithm for its structure.
  • The paper's measurements show total runtime speedups of up to 13.54x on quadratic programs, 21.02x on linear programs, and 1.67x on second-order cone programs, with backward gradient cosine similarity 0.992 on QP against a high-precision KKT solve.
  • The method scales to large sparse problems (up to 5000x2000 in the experiments), where the paper reports that competing differentiable layers fail to generate results.

Reading between the lines

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

  • Inference: the reported speedups should grow with the fraction of inactive constraints, since the backward QP keeps only the active rows; the paper does not isolate this dependence, but it follows directly from the reformulation.
  • Inference: because the active set is read off a finite-tolerance solver's output, the natural stress test is a degenerate problem where a dual multiplier is near zero; Theorem 1 is stated without explicitly requiring strictly positive multipliers on all active inequalities, and this is where the backward QP gradient could diverge from the exact KKT value.
  • Inference: the same active-set-to-QP trick could be applied to higher-order derivative computations, such as Hessian-vector products, or to non-convex forward maps near a local minimum, since the derivation only uses the KKT structure at the returned point; the paper hints at the non-convex extension but does not test it.
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

4 major / 5 minor

Summary. The paper proposes BPQP, a differentiable convex optimization layer that reformulates the backward pass of implicit differentiation as an equality-constrained quadratic program (Eq. 6) whose KKT system coincides with the linear system obtained by differentiating the KKT conditions of the forward problem (Eq. 3), under an active-set replacement. The backward QP is then solved with OSQP, decoupling the forward and backward solvers. Experiments compare runtime and gradient accuracy against CVXPY, qpth/OptNet, Alt-Diff, JAXOpt, and an Exact matrix-inverse baseline on randomly generated QPs, LPs, and SOCPs, and on a CSI 500 portfolio-optimization task, reporting large speedups and competitive or improved decision metrics.

Significance. If the equivalence in Theorem 1 is established under the right assumptions, the central idea is practically useful: it turns the backward linear solve into a convex QP that can leverage mature first-order solvers, and the decoupling of forward and backward passes is attractive for large-scale end-to-end learning. The paper provides a proof in Appendix A.2, releases code through Qlib, and reports extensive runtime and accuracy experiments. Its value, however, depends on the exactness of the backward gradient under the stated conditions, which are currently too weak, and on the credibility of the LP and SOCP experimental claims. The theoretical contribution is a reformulation with empirical support rather than a new mathematical theory.

major comments (4)
  1. [Theorem 1 / Appendix A.2] The proof of Theorem 1 replaces the complementarity block D(λ*)G(z*) and D(g(z*)) in Eq. (3) with the equality block G_+ in Eqs. (20)-(21). This replacement is exact only when the active set is known exactly and every active inequality satisfies λ*_i > 0 (strict complementarity). The stated assumptions—'not primal infeasible' and 'the corresponding Jacobian vector ∇yL exists'—do not imply strict complementarity. When a constraint is weakly active (g_i(z*)=0, λ*_i=0), the i-th complementarity row of Eq. (3) is identically 0=0 and imposes no constraint on \tilde z, while Eq. (6) enforces G_{i+}\tilde z=0; the two systems can then produce different gradients. A concrete degenerate case is min_z 0.5||z||^2 + q^T z subject to z_i ≤ 0 at q=0 with loss L=z_1+z_2, where Eq. (3) admits both \tilde z=(-1,-1) and \tilde z=(0,0), but Eq. (6) forces \tilde z=0. In addition, OSQP returns solutions only to finite tolerance (Section 4.2), so the active set recovered from the forward pass is approximate; a near-active constraint that is misclassified changes Eq. (6) and hence the gradient. Theorem 1 should state the strict-complementarity and exact-active-set assumptions, or be reformulated as an approximate-gradient statement with a perturbation bound.
  2. [Section 5.1, Eq. (17)] The experiments labeled 'LP' solve minimize θ^T z + ε||z||^2_2 subject to Az=b, Gz≤h with ε=10^{-6}. This is a strictly convex quadratic program, not a linear program. For a genuine LP (P=0), the KKT matrix in Eq. (3) can be singular and the backward solution nonunique; the paper provides no analysis of this case. Consequently, the claimed 21.02× speedup on LP and the statement that BPQP handles LP layers are not demonstrated for standard LPs. The experiment should be reported as 'regularized LP' or, preferably, true LP instances should be tested with a discussion of degeneracy and uniqueness.
  3. [Table 1] Table 1 gives absolute times with row labels '(scale 1.0e-04)' for Exact, CVXPY, qpth/OptNet, Alt-Diff, and JAXOpt, but '(scale 1.0e+00)' for BPQP. As printed, the BPQP entries are in different units from the other entries, and the quoted speedups in the abstract and Section 1 are not recoverable unless all rows are read with the same implicit scale. For example, the QP 100×20 total time for Exact is 484.2×10^{-4} s and for BPQP is 35.1×10^{-4} s if the same scale is used, giving the ~13.5× speedup; with the printed scale labels, BPQP would appear to be orders of magnitude slower. Please restate the table with one explicit unit for all rows or report raw times in a common unit.
  4. [Section 6 and 'General Gradients' in Section 4.1] The claim that BPQP 'is still equipped to reformulate the backward pass as a QP' for non-convex problems is not supported. For a non-convex objective, the matrix P' in Eq. (6) need not be positive semidefinite, so the backward problem is not a convex QP and OSQP's convergence guarantees (Section 4.2) do not apply. The 'General Gradients' paragraph defines gradients that preserve the KKT norm at intermediate iterates, but Appendix A.4 does not show that these surrogate vectors equal or approximate the true loss gradient; it only states an identity dr(k)=0 and concludes norm preservation. This is a heuristic, not a theorem. Either provide conditions under which the KKT-norm-preserving update yields the correct gradient, or remove the non-convex claims from the discussion.
minor comments (5)
  1. [Definition 1 and Section 3.2] There are dimension errors in the notation: h should map R^d to R^m and g should map R^d to R^n, and the backward variables in Eq. (3) should be \tilde λ ∈ R^n and \tilde ν ∈ R^m, not the reverse as printed.
  2. [Section 4.1] The condition list for equivalence of Eq. (5) and Eq. (3) is confusing: D(\tilde λ)G' = D(λ*)G(z*) is a matrix equation involving the unknown \tilde λ, not a condition that can be checked a priori. The exposition would benefit from stating that the active-set replacement is used precisely to avoid this issue.
  3. [Appendix A.2, Eq. (19)] There is a typo: 'the original optimization problem in can be reformulated' should read 'in Definition 1 can be reformulated'.
  4. [Table 3] The text says 'The CosSim. of all methods are small enough for SOCP' but the table shows values of 1.00; the intended wording is probably 'close to one' or 'sufficiently high'.
  5. [Section 6] The sentence 'While its hard to perform experiments on non-convex problem due to the lack of baselines' contains grammar errors and should be rephrased.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: BPQP's backward QP reformulation is a mathematical equivalence, not a refit or self-citation reduction.

full rationale

The load-bearing step is Theorem 1 and Appendix A.2: Eq. (21) is obtained by replacing the complementarity block D(lambda*)G(z*) with G+(z*) after restricting to the active set, and Eq. (22) is exactly the KKT system of the equality-constrained QP (6). This is an algebraic equivalence between two linear systems, with no fitted parameter entering the gradient formula. P', A', G'_+, q', c'_+, and b' are taken directly from the forward-pass KKT data and the given loss gradient; no term is tuned to reproduce the target gradient. The added regularizers delta and epsilon are fixed constants (10^-6) and are not calibrated against the quantities being predicted. The QP/LP/SOCP speedups and portfolio results are benchmarked against external baselines (OptNet, CVXPY, JAXOpt, Alt-Diff, Exact, Two-Stage), so they do not reduce to the paper's own assumptions. The active-set replacement in Eq. (20) can fail under weak complementarity or finite OSQP tolerance, but that is a correctness/robustness concern about the assumption, not a circularity: the derivation does not define the gradient in terms of itself. The 'General Gradients' paragraph is under-specified but is not used to fit any parameter or to produce the reported results. No load-bearing self-citation was found; [24] is a code release and [14,16] are external prior work. Therefore the derivation is self-contained with respect to circularity.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central derivation uses standard KKT and IFT mathematics plus an active-set equality reformulation. The main added assumptions are the unstated constraint qualification, strict complementarity, and convergence of the OSQP iterative refinement. The regularization constants epsilon and delta are chosen to make the problems differentiable and invertible, and the loss weight beta is a hand-chosen hyperparameter.

free parameters (5)
  • LP regularization epsilon = 1e-6
    Added to Eq. (17) to make the LP's optimal solution differentiable; without it P(z*) = 0 and the backward KKT system is singular. This means the LP benchmark is a regularized LP, not a true LP.
  • KKT regularization delta = 1e-6
    Added to the backward KKT matrix in Eq. (10) to guarantee invertibility for redundant constraints; the gradient is computed for a perturbed system and recovered via iterative refinement.
  • OSQP termination tolerances = abs/rel tolerance 1e-3, max iterations 4000 (simulated); 1e-5 (portfolio)
    Termination criteria for the forward and backward QP solvers; influence both speed and gradient accuracy in all experiments.
  • Portfolio loss weight beta = 0.1
    Weight balancing decision regret and prediction error in the portfolio optimization loss (Appendix A.6); chosen by hand, not fitted.
  • Risk aversion coefficient gamma = 1
    Fixed risk aversion in the mean-variance optimization problem Eq. (18); affects portfolio weights and the reported Sharpe ratios.
assumptions (4)
  • domain assumption The convex optimization layer satisfies KKT necessity (a constraint qualification such as Slater or LICQ) and the KKT matrix is invertible at the optimum.
    Theorem 1 and Eq. (3) require the Jacobian of the KKT conditions to be invertible; this condition is not stated in Definition 1.
  • ad hoc to paper The active set is correctly identified from the forward solution and strict complementarity holds for active constraints, so replacing inequalities with equalities in Eq. (20) preserves the linear system solution.
    The proof in Appendix A.2 uses the active equality reformulation; weakly active constraints (g_i = 0 and lambda_i = 0) would make the reduction inexact. The paper does not discuss this case.
  • standard math OSQP's ADMM with iterative refinement converges to the exact solution of the regularized KKT system in practice, with one backward and one forward solve as in Eq. (11).
    The efficiency claim depends on OSQP's convergence behavior, which is cited from [16] but not proven here.
  • domain assumption The Hessian of the Lagrangian P(z*,nu*,lambda*) is positive semidefinite for the problem classes considered, so the backward QP is convex.
    Theorem 1 requires P' to be positive semidefinite; this holds for convex f and g with lambda* >= 0, but it is not stated as an explicit condition.

how reviews work

0 comments
Cite this review

Pith. "Pith review of BPQP: A Differentiable Convex Optimization Framework for Efficient End-to-End Learning." pith.science (2026). https://pith.science/paper/HRNCYU3O

@misc{pith2026241119285,
  author       = {Pith},
  title        = {Pith review of: BPQP: A Differentiable Convex Optimization Framework for Efficient End-to-End Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HRNCYU3O}},
  note         = {Machine review of arXiv:2411.19285}
}
read the original abstract

Data-driven decision-making processes increasingly utilize end-to-end learnable deep neural networks to render final decisions. Sometimes, the output of the forward functions in certain layers is determined by the solutions to mathematical optimization problems, leading to the emergence of differentiable optimization layers that permit gradient back-propagation. However, real-world scenarios often involve large-scale datasets and numerous constraints, presenting significant challenges. Current methods for differentiating optimization problems typically rely on implicit differentiation, which necessitates costly computations on the Jacobian matrices, resulting in low efficiency. In this paper, we introduce BPQP, a differentiable convex optimization framework designed for efficient end-to-end learning. To enhance efficiency, we reformulate the backward pass as a simplified and decoupled quadratic programming problem by leveraging the structural properties of the KKT matrix. This reformulation enables the use of first-order optimization algorithms in calculating the backward pass gradients, allowing our framework to potentially utilize any state-of-the-art solver. As solver technologies evolve, BPQP can continuously adapt and improve its efficiency. Extensive experiments on both simulated and real-world datasets demonstrate that BPQP achieves a significant improvement in efficiency--typically an order of magnitude faster in overall execution time compared to other differentiable optimization layers. Our results not only highlight the efficiency gains of BPQP but also underscore its superiority over differentiable optimization layer baselines.

Figures

Figures reproduced from arXiv: 2411.19285 by the authors.

Figure 1
Figure 1. The learning process of BPQP: the previous layer outputs [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. Sensitivity analysis under 500×100 set￾ting. 5.2 Real-world End-to-End Portfolio Optimization Portfolio optimization is a fundamental problem for asset allocation in finance. It involves constructing and balancing the investment portfolio periodically to maximize profit and minimize risk. The problem is an important use case of end-to-end learning and can also be solved utilizing differentiable convex optimization l… view at source ↗
Figure 4
Figure 4. The prediction and decision error/loss of methods with different objectives [PITH_FULL_IMAGE:figures/full_fig_p018_4.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 32 canonical work pages

  1. [1]

    Differentiable convex optimization layers

    Akshay Agrawal, Brandon Amos, Shane Barratt, Stephen Boyd, Steven Diamond, and J Zico Kolter. Differentiable convex optimization layers. Advances in neural information processing 10 systems, 32, 2019

    work page 2019

  2. [2]

    Deep declarative networks

    Stephen Gould, Richard Hartley, and Dylan Campbell. Deep declarative networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(8):3988–4004, 2021

    work page 2021

  3. [3]

    A tutorial on energy- based learning

    Yann LeCun, Sumit Chopra, Raia Hadsell, M Ranzato, and Fujie Huang. A tutorial on energy- based learning. Predicting structured data, 1(0), 2006

    2006

  4. [4]

    Generic methods for optimization-based modeling

    Justin Domke. Generic methods for optimization-based modeling. In Artificial Intelligence and Statistics, pages 318–326. PMLR, 2012

    work page 2012

  5. [5]

    Smart predict-and-optimize for hard combina- torial optimization problems

    Jayanta Mandi, Peter J Stuckey, Tias Guns, et al. Smart predict-and-optimize for hard combina- torial optimization problems. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pages 1603–1610, 2020

    work page 2020

  6. [6]

    predict, then optimize

    Adam N Elmachtoub and Paul Grigas. Smart “predict, then optimize”. Management Science, 68(1):9–26, 2022

    2022

  7. [7]

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

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

    work page 2019

  8. [8]

    Stochastic distribution control system design: a convex optimization approach

    Lei Guo and Hong Wang. Stochastic distribution control system design: a convex optimization approach. Springer, 2010

    work page 2010

Show all 44 references
  1. [9]

    Real-time convex optimization in signal processing

    John Mattingley and Stephen Boyd. Real-time convex optimization in signal processing. IEEE Signal Processing Magazine, 27(3):50–61, 2010

    2010

  2. [10]

    Efficient and modular implicit differentiation

    Mathieu Blondel, Quentin Berthet, Marco Cuturi, Roy Frostig, Stephan Hoyer, Felipe Llinares- López, Fabian Pedregosa, and Jean-Philippe Vert. Efficient and modular implicit differentiation. arXiv preprint arXiv:2105.15183, 2021

    2021 arXiv

  3. [11]

    Efficient multiple hyperparameter learning for log-linear models

    Chuan-sheng Foo, Andrew Ng, et al. Efficient multiple hyperparameter learning for log-linear models. In Advances in neural information processing systems, volume 20, 2007

    2007

  4. [12]

    Alternating differentiation for optimization layers

    Haixiang Sun, Ye Shi, Jingya Wang, Hoang Duong Tuan, H Vincent Poor, and Dacheng Tao. Alternating differentiation for optimization layers. In The Eleventh International Conference on Learning Representations, 2022

    2022

  5. [13]

    An implicit function theorem

    K Jittorntrum. An implicit function theorem. Journal of Optimization Theory and Applications, 25(4):575–577, 1978

    1978

  6. [14]

    Optnet: Differentiable optimization as a layer in neural networks

    Brandon Amos and J Zico Kolter. Optnet: Differentiable optimization as a layer in neural networks. In International Conference on Machine Learning, pages 136–145. PMLR, 2017

    2017

  7. [15]

    Differenti- ating through a cone program

    Akshay Agrawal, Shane Barratt, Stephen Boyd, Enzo Busseti, and Walaa M Moursi. Differenti- ating through a cone program. arXiv preprint arXiv:1904.09043, 2019

    1904 arXiv

  8. [16]

    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(4):637–672, 2020

    2020

  9. [17]

    The simplex method for quadratic programming

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

    1959

  10. [18]

    Efficient differentiable quadratic programming layers: an admm approach

    Andrew Butler and Roy H Kwon. Efficient differentiable quadratic programming layers: an admm approach. Computational Optimization and Applications, 84(2):449–476, 2023

    2023

  11. [19]

    Cvxpy: A python-embedded modeling language for convex optimization

    Steven Diamond and Stephen Boyd. Cvxpy: A python-embedded modeling language for convex optimization. The Journal of Machine Learning Research, 17(1):2909–2913, 2016

    2016

  12. [20]

    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, 169:1042–1068, 2016. 11

    2016

  13. [21]

    Mipaal: Mixed integer program as a layer

    Aaron Ferber, Bryan Wilder, Bistra Dilkina, and Milind Tambe. Mipaal: Mixed integer program as a layer. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pages 1504–1511, 2020

    2020

  14. [22]

    Implicit mle: backpropagating through discrete exponential family distributions

    Mathias Niepert, Pasquale Minervini, and Luca Franceschi. Implicit mle: backpropagating through discrete exponential family distributions. InAdvances in Neural Information Processing Systems, 2021

    2021

  15. [23]

    Satnet: Bridging deep learning and logical reasoning using a differentiable satisfiability solver

    Po-Wei Wang, Priya Donti, Bryan Wilder, and Zico Kolter. Satnet: Bridging deep learning and logical reasoning using a differentiable satisfiability solver. In International Conference on Machine Learning, pages 6545–6554. PMLR, 2019

    2019

  16. [24]

    Qlib: An ai-oriented quantitative investment platform

    Xiao Yang, Weiqing Liu, Dong Zhou, Jiang Bian, and Tie-Yan Liu. Qlib: An ai-oriented quantitative investment platform. arXiv preprint arXiv:2009.11189, 2020

    2009 arXiv

  17. [25]

    Robust linear programming discrimination of two linearly inseparable sets

    Kristin P Bennett and Olvi L Mangasarian. Robust linear programming discrimination of two linearly inseparable sets. Optimization methods and software, 1(1):23–34, 1992

    1992

  18. [26]

    End-to-end risk budgeting portfolio optimiza- tion with neural networks

    A Sinem Uysal, Xiaoyue Li, and John M Mulvey. End-to-end risk budgeting portfolio optimiza- tion with neural networks. Annals of Operations Research, pages 1–30, 2023

    2023

  19. [27]

    H. M. Markowitz. Portfolio selection. The journal of finance, 7(1):77–91, 1952

    1952

  20. [28]

    V olatility clustering in financial markets: a microsimulation of interacting agents

    Thomas Lux and Michele Marchesi. V olatility clustering in financial markets: a microsimulation of interacting agents. International journal of theoretical and applied finance, 3(04):675–702, 2000

    2000

  21. [29]

    Improved svrg for non-strongly-convex or sum-of-non- convex objectives

    Zeyuan Allen-Zhu and Yang Yuan. Improved svrg for non-strongly-convex or sum-of-non- convex objectives. In International conference on machine learning, pages 1080–1089. PMLR, 2016

    2016

  22. [30]

    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. arXiv preprint arXiv:2104.12225, 2021

    2021 arXiv

  23. [31]

    End-to-end learning for optimization via constraint-enforcing approximators

    Rares Cristian, Pavithra Harsha, Georgia Perakis, Brian L Quanz, and Ioannis Spantidakis. End-to-end learning for optimization via constraint-enforcing approximators. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 7253–7260, 2023

    2023

  24. [32]

    End-to-end stochastic optimization with energy-based model

    Lingkai Kong, Jiaming Cui, Yuchen Zhuang, Rui Feng, B Aditya Prakash, and Chao Zhang. End-to-end stochastic optimization with energy-based model. Advances in Neural Information Processing Systems, 35:11341–11354, 2022

    2022

  25. [33]

    Learning the travelling salesperson problem requires rethinking generalization

    Chaitanya K Joshi, Quentin Cappart, Louis-Martin Rousseau, and Thomas Laurent. Learning the travelling salesperson problem requires rethinking generalization. Constraints, pages 1–29, 2022

    2022

  26. [34]

    Learning combinatorial optimization algorithms over graphs

    Elias Khalil, Hanjun Dai, Yuyu Zhang, Bistra Dilkina, and Le Song. Learning combinatorial optimization algorithms over graphs. Advances in neural information processing systems, 30, 2017

    2017

  27. [35]

    Combinatorial opti- mization by graph pointer networks and hierarchical reinforcement learning

    Qiang Ma, Suwen Ge, Danyang He, Darshan Thaker, and Iddo Drori. Combinatorial opti- mization by graph pointer networks and hierarchical reinforcement learning. arXiv preprint arXiv:1911.04936, 2019

    1911 arXiv

  28. [36]

    Attention, learn to solve routing problems! arXiv preprint arXiv:1803.08475, 2018

    Wouter Kool, Herke Van Hoof, and Max Welling. Attention, learn to solve routing problems! arXiv preprint arXiv:1803.08475, 2018

    2018 arXiv

  29. [37]

    End-to-end risk budgeting portfolio optimization with neural networks

    Ayse Sinem Uysal, Xiaoyue Li, and John M Mulvey. End-to-end risk budgeting portfolio optimization with neural networks. arXiv preprint arXiv:2107.04636, 2021

    2021 arXiv

  30. [38]

    Decision-focused learning without differentiable optimization: Learning locally optimized decision losses

    Sanket Shah, Kai Wang, Bryan Wilder, Andrew Perrault, and Milind Tambe. Decision-focused learning without differentiable optimization: Learning locally optimized decision losses. arXiv preprint arXiv:2203.16067, 2022. 12

    2022 arXiv

  31. [39]

    Automatically learning compact quality-aware surrogates for optimization problems

    Kai Wang, Bryan Wilder, Andrew Perrault, and Milind Tambe. Automatically learning compact quality-aware surrogates for optimization problems. Advances in Neural Information Processing Systems, 33:9586–9596, 2020

    2020

  32. [40]

    Surco: Learning linear surrogates for combinatorial nonlinear optimization problems

    Aaron M Ferber, Taoan Huang, Daochen Zha, Martin Schubert, Benoit Steiner, Bistra Dilkina, and Yuandong Tian. Surco: Learning linear surrogates for combinatorial nonlinear optimization problems. In International Conference on Machine Learning , pages 10034–10052. PMLR, 2023

    2023

  33. [41]

    Landscape surrogate: Learning decision losses for mathematical optimization under partial information

    Arman Zharmagambetov, Brandon Amos, Aaron Ferber, Taoan Huang, Bistra Dilkina, and Yuandong Tian. Landscape surrogate: Learning decision losses for mathematical optimization under partial information. Advances in Neural Information Processing Systems, 36, 2024

    2024

  34. [42]

    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(3):1042–1068, June 2016

    2016

  35. [43]

    Operator splitting for a homogeneous embedding of the linear comple- mentarity problem

    Brendan O’Donoghue. Operator splitting for a homogeneous embedding of the linear comple- mentarity problem. SIAM Journal on Optimization, 31:1999–2023, August 2021

    1999

  36. [44]

    Comparing technical and fun- damental indicators in stock price forecasting

    Erhan Beyaz, Firat Tekiner, Xiao-jun Zeng, and John Keane. Comparing technical and fun- damental indicators in stock price forecasting. In 2018 IEEE 20th International Confer- ence on High Performance Computing and Communications; IEEE 16th International Con- ference on Smart ...

    2018

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.