Pith. sign in

REVIEW 2 major objections 5 minor 92 references

Smooth Learning with Hard Constraints via Legendre-Regularized Policies

T0 review · 2 major / 5 minor · reviewed 2026-07-31 · deepseek-v4-flash

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

Pith's one-line read A regularized optimization wrapper turns hard-constrained decision policies into smooth, differentiable maps without losing expressive power.

desk verdict Solid theory with a narrower practical reach than claimed; worth peer review, with experiments needing a projection baseline and significance testing. read the letter →

arxiv 2607.24007 v1 pith:VYINIHFX submitted 2026-07-27 math.OC cs.LG

Zikun Lin , Rui Chen , Yijie Wang This is my paper · ORCID
classification math.OCcs.LG MSC 90C2590C15
keywords decision-focusedlearningcontextualoptimizationLegendre-regularizedpolicieshardconstraintsdifferentiablelayersuniversalapproximationregularizedoptimizersprescriptiveanalytics
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

Decision-focused learning faces a three-way tension: policies must be expressive, must respect hard feasibility constraints, and must be smooth enough for gradient training. This paper proposes Legendre-regularized policies, where a learned latent vector is fed into a regularized optimization problem over the original feasible region and the solution is the decision. The paper establishes that, when the regularizer is Legendre-type (strongly convex, essentially smooth, with twice-differentiable conjugate), the resulting solution map is single-valued, maps onto the relative interior of the feasible set, has an explicit Jacobian, is Lipschitz continuous, and can be made arbitrarily smooth. It also proves that these Legendre conditions are necessary for those properties, and that composing the map with a universal approximator yields uniform approximation of any continuous feasible policy on compact contexts. If correct, this resolves the trilemma and unifies explicit regularizers like log-barrier and entropy with perturbation-based differentiable optimizers; experiments on newsvendor and resource allocation report lower decision costs than the benchmarks.

What carries the argument

The central object is the admissible Legendre regularizer φ (Definition 1): a proper lower-semicontinuous convex function whose effective domain has relative interior equal to S°, and whose reduction ψ(y) = φ(w̄ + Ny) on the affine hull is μ-strongly convex, essentially smooth, and has twice-differentiable convex conjugate. The workhorse identity is w_{F,φ}(z) = w̄ + N∇ψ*(−Bz), derived from the first-order condition Bz + ∇ψ(y) = 0; it converts the optimization into the convex-analytic inversion ∇ψ* = (∇ψ)^{-1}. The Jacobian −N∇²ψ*(−Bz)B is the Hessian of the conjugate, which explains why smoothness of the conjugate transfers to the policy and why the Lipschitz constant is inherited from the

What would settle it

Look for any convex feasible set S and non-Legendre regularizer whose solution map is single-valued, everywhere differentiable, Lipschitz, and bijective onto the relative interior; Proposition 1 says none exists. The natural suspect, a quadratic penalty on a box (Euclidean projection), fails at boundary saturation and differentiability, so a valid counterexample would immediately refute the necessity result.

Watch

Extended reading notes

Core claim

At the center is the optimizer map w_{F,φ}(z) = argmin_{w∈S} {⟨Fz,w⟩ + φ(w)}. The paper claims that under Definition 1 (admissible Legendre regularizer) this map is a single-valued function from R^k onto the relative interior S°, with the explicit form w = w̄ + N ∇ψ*(−Bz) and Jacobian −N ∇²ψ*(−Bz)B, and it is Lipschitz with constant bounded by ||F||/μ. The same Legendre assumptions are shown necessary: any regularized optimizer that is uniquely defined, bijective onto the interior, Lipschitz, and everywhere differentiable forces the regularizer's reduction to be strongly convex, essentially smooth, and have a twice-differentiable conjugate. The closing theoretical step is a universal approxi

Load-bearing premise

Everything rests on being able to construct an admissible Legendre regularizer — a μ-strongly convex, essentially smooth function with twice-differentiable conjugate — for the specific feasible set, and to evaluate its solution map and Jacobian cheaply; the paper demonstrates such regularizers for boxes and grouped-budget polytopes but gives no general recipe for arbitrary convex feasible sets.

Editorial extensions

If this is right

  • Because smoothness lives in the decision map rather than a surrogate loss, Legendre-regularized policies can be trained on any subdifferentiable downstream loss, including nonlinear and piecewise-linear costs.
  • Policies are feasible by construction at every training step; no projection or repair layer is needed, and gradients exist everywhere rather than only almost everywhere.
  • Universal approximation transfers from the latent model to the policy class, so hard feasibility and smoothness do not cap expressive power.
  • Log-barrier, entropic, and perturbation-based smooth optimizers are instances of one construction, giving a unified view of explicit and implicit smoothing.
  • With smoother regularizers the policy becomes C^{k-1}, which enables sensitivity analysis and higher-order training methods, not just first-order backpropagation.

Reading between the lines

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

  • Editorial extension: the necessity result suggests a design checklist for any smooth-optimizer layer: if the layer must be single-valued, everywhere differentiable, bijective onto the feasible interior, and Lipschitz, its regularizer must be of Legendre type; quadratic penalties fail the essential-smoothness test.
  • The bijection between latent space and the relative interior points to a regularizer-induced coordinate system on the feasible region, which could be exploited for uncertainty quantification or calibrated decision sets.
  • A testable direction the paper leaves open is scaling: for general polyhedral feasible sets with no closed-form layer, a Newton-iteration layer with implicit differentiation would need to be benchmarked against the closed-form box and polytope cases to see whether the theoretical guarantees survive finite iterations.
  • Because the approximation theorem only needs a dense latent model, any dense function class (not only neural networks) could replace the predictor and retain the same uniform approximation guarantee.
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

2 major / 5 minor

Summary. The paper introduces Legendre-regularized policies (LRPs) for contextual decision-making under hard feasibility constraints. A policy is defined as π_g(x)=w_{F,φ}(g(x)), where w_{F,φ}(z) is the minimizer of ⟨Fz,w⟩+φ(w) over the feasible set S and φ is an 'admissible Legendre regularizer.' The main theoretical results are: Theorem 1 gives single-valuedness, containment in the relative interior, an explicit Jacobian, Lipschitz continuity, and surjectivity; Proposition 1 establishes a converse necessity; Proposition 2 proves uniform universal approximation on compact contexts. The paper also identifies log-barrier, entropic, and Gaussian-perturbation regularizers as instances and reports experiments on a newsvendor problem and two resource-allocation problems with box or group-budget constraints.

Significance. The theoretical framework is elegant and largely correct. The affine-hull reduction and the duality expression w = w̄ + N∇ψ*(-Bz) make the policy analytically tractable, and the explicit Jacobian is a concrete contribution for end-to-end training. The universality result is reassuring: smoothing does not come at the cost of expressiveness. The necessity result cleanly distinguishes Legendre regularization from projection-based or quadratic smoothing. The experiments are honest, including settings where the SAA benchmark wins, and the improvements over PADR/DD-SAA are meaningful in the tested regimes. The main caveat is that the computational premise—availability and efficient evaluation of an admissible regularizer—is only demonstrated for separable boxes and one low-dimensional coupled polytope.

major comments (2)
  1. [§2.1, Definition 1; §2.2, Eq. (2)] The proof of Theorem 1 and the sentence 'Since φ has its effective domain contained in S' rely on a domain condition that is not stated in Definition 1. Definition 1 only says S°=(dom φ)°. For a closed convex S with nonempty relative interior, equality of relative interiors together with convexity does imply dom φ⊆S, but this implication is not shown; and if '°' is read as ordinary interior (as in the full-dimensional discussion), the condition is incompatible with lower-dimensional S. Please state explicitly that dom φ⊆S (or cl(dom φ)=S) and clarify that '°' denotes relative interior throughout. This is load-bearing because without such a condition the minimization in (2) is not restricted to S, so feasibility by construction is not guaranteed.
  2. [§3.3–3.4, §A.4, §4] The practical claim that LRPs are broadly trainable for hard-constrained problems goes beyond the evidence. All experiments use either coordinatewise separable boxes, where the maps are closed-form, or one 20-dimensional coupled set with two group-budget inequalities, where the dual solve is two-dimensional. For a general polyhedral S with many coupled inequalities, evaluating w_{F,φ} requires a convex optimization of the same scale as the original decision problem, and differentiating it requires an additional KKT linear solve; no complexity or numerical-stability analysis is given, and §4 lists 'scalable algorithms' as an open direction. This does not invalidate Theorem 1 or Proposition 2, but it is a load-bearing gap for the paper's practical motivation. I recommend either adding a larger coupled experiment or explicitly narrowing the scope of the practical claim and stating the compu
minor comments (5)
  1. [§2.1, Definition 1] The notation S° and (dom φ)° should be defined as relative interiors in one place; the current text alternates between 'relative interior' and 'interior' in the full-dimensional discussion.
  2. [§2.3.3] The claim that under the assumptions of Berthet et al. (2020) φ_ptb is strongly convex, Legendre-type, and has twice-differentiable conjugate would benefit from a precise reference (e.g., proposition number) and a short argument, since not every twice-differentiable expected maximum has a strongly convex conjugate.
  3. [§3.1] The projection step bπ_a(x) used for non-LRP baselines is mentioned but not described in full (e.g., the projection problem is assumed to be cheap); state whether this projection is computed exactly for all methods.
  4. [Tables 1–3] Standard deviations are reported but no significance tests or paired comparisons. In several rows the LRP advantage over PADR is modest relative to spread; a brief statement on statistical significance would improve interpretation.
  5. [Appendix A.4] For the group-budget LRP-Log and LRP-Ent implementations, the derivation of the dual Newton iteration and the differentiation of the KKT system is sketched only verbally; a short pseudocode or derivation would help reproducibility.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the core theorems are derived from explicit technical assumptions using standard external results; the only self-citations are related-work mentions and are not load-bearing.

full rationale

The paper's central claims are mathematical derivations from explicit premises. Theorem 1 (Section 2.4) takes Definition 1's admissibility conditions — strong convexity, essential smoothness, and twice-differentiability of the conjugate — and derives the solution map's uniqueness, relative-interior image, Jacobian formula, Lipschitz constant, and higher-order smoothness via standard convex-analysis results (Rockafellar 1970; Rockafellar & Wets 2009). The conditions are stated on the reduced regularizer ψ, not directly on the desired solution-map properties, so Theorem 1 is not a tautology; it supplies the bridge. Proposition 1 is a genuine converse, proving that if a regularized optimizer has the target regularity properties, then the Legendre-type conditions must hold. Proposition 2 (Section 2.5) is a constructive universality argument: it shrinks any continuous feasible policy into the relative interior, builds a continuous latent function h via ∇ψ, and then invokes the universal-approximation assumption on the model class G. This is a proof by construction, not a fitted-input-called-prediction step. The only self-citations (Wang et al. 2026a, 2026b) occur in related-work contexts and in a survey of quadratic smoothing; they do not support any load-bearing premise. The authors' own Section 4 limitation — that scalable algorithms for evaluating/differentiating these layers in general large-scale problems remain open — is a scope and computational-tractability risk, not a circularity. All experiments use held-out confirmation replications and compare against external benchmarks, so there is no statistical reduction of a prediction to a fitted value. Overall, the derivation is self-contained; the score of 1 reflects the presence of minor, non-load-bearing self-citations rather than any circular structure.

Assumptions & free parameters 4 free parameters · 6 assumptions · 1 invented entities

The central theorems carry no fitted parameters: w̄ and δ in Prop 2 are proof devices with explicit bounds (δM ≤ ε/2), and the choice of w̄ does not affect the policy map w_{F,φ}(z), which is defined over R^n independently of the reduced representation. The listed free parameters are hyperparameters of the experimental method, selected on separate tuning replications, and do not feed the theoretical claims. The axioms are standard convex-analysis theorems plus the explicit domain assumptions of §2; the only borrowed condition is the admissibility of φ_ptb from Berthet et al. (2020).

free parameters (4)
  • smoothing schedule (τ_0, τ_min, δ, s) plus learning rate/weight decay = tuned per setting via Bayesian optimization on 5 tuning replications (App. A.2)
    Controls the scale and annealing of the regularizer in the experiments; the LRP map and test costs depend on it. It is a method hyperparameter, not part of the theoretical claims.
  • regularizer choice φ ∈ {log, ent, ptb} = all three tested; best variant varies by setting
    Any admissible φ yields the Theorem 1 guarantees; which one performs best is empirically selected. An experimental degree of freedom, not a fitted constant in the theory.
  • latent map F (with B = N^T F full row rank) = F = I_n in all experiments
    Design choice inside the policy class; theory requires only rank(B)=r. In experiments the latent dimension equals the decision dimension.
  • MLP architecture (3 hidden layers, widths 16-32-16, ReLU) = fixed as in §3.1
    Used to realize G; Assumption 1 needs G to be a universal approximator in the limit, which any standard MLP satisfies asymptotically. Architecture is a validation-set choice.
assumptions (6)
  • standard math Legendre-type duality theorems (Rockafellar 1970 Thm 25.5, 26.5, Cor 26.3.1; Rockafellar-Wets 2009 Prop 12.60): for μ-strongly convex essentially smooth ψ, ψ* is differentiable, ∇ψ* is (1/μ)-Lipschitz and inverse to ∇ψ, and Lipschitz gradients are dual to strong convexity.
    Invoked directly in Lemma 1 and in the proofs of Theorem 1 and Proposition 1 (§2.2, §2.4).
  • standard math Envelope/Danskin theorem for extended-real-valued convex functions (Bertsekas 1971 Prop A.22), used to identify ∇ψ* with the argmin map of the regularized problem.
    Core step in Proposition 1 (§2.4): differentiates the optimal value of problem (6) and identifies the gradient of ψ*.
  • domain assumption S is a nonempty closed convex set with aff(S) = {w : Aw = b}, A ∈ R^{p×n} full row rank, and an orthonormal basis N of ker(A); the reduced region Y = {y : w̄ + Ny ∈ S°} is open convex with int(dom ψ) = Y.
    Assumed at the start of §2 for all theory; the experiments use polytopes of this form.
  • domain assumption Assumption 1: the latent model class G can uniformly approximate any continuous h : X → R^k on compact X.
    Used in Proposition 2; standard universal-approximation hypothesis on the latent network class.
  • domain assumption For LRP-Ptb: under the conditions in Berthet et al. (2020), the implicit regularizer φ_ptb (convex conjugate of the expected maximum) is strongly convex, of Legendre type, and has a twice-differentiable conjugate.
    §2.3.3 asserts admissibility via citation rather than reproving it; load-bearing for the LRP-Ptb variant's place in the framework.
  • standard math Inverse function theorem and standard composition/differentiability results for C^k maps.
    Used for Lemma 1 property 4 and the higher-order smoothness statements in Theorem 1.
invented entities (1)
  • Admissible Legendre regularizer (Definition 1) and the LRP policy class Π_{F,φ}(G) independent evidence
    purpose: Construction that yields feasible, smooth, surjective-onto-S° policies with explicit Jacobians for decision-focused learning.
    A definition, not an unexplained postulate: it is instantiated by explicit regularizers (log barrier §2.3.1, entropy §2.3.2, Gaussian-perturbed oracle §2.3.3) with closed-form maps in the experiments (§A.4), and justified ex post by the necessity theorem (Prop 1). No new physical or empirical entity is introduced.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Smooth Learning with Hard Constraints via Legendre-Regularized Policies." pith.science (2026). https://pith.science/paper/VYINIHFX

@misc{pith2026260724007,
  author       = {Pith},
  title        = {Pith review of: Smooth Learning with Hard Constraints via Legendre-Regularized Policies},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VYINIHFX}},
  note         = {Machine review of arXiv:2607.24007}
}
read the original abstract

We revisit contextual optimization from the perspective of policy class design. A desirable policy class should be expressive enough to learn rich context-decision relationships, should enforce hard feasibility constraints rather than soft penalty terms, and should remain smooth enough for gradient-based training on downstream decision losses. Existing approaches usually emphasize only part of these requirements. We propose Legendre-regularized policies, which parameterize decisions as solutions of regularized optimization problems over the original feasible region. This construction yields policies that are feasible by construction and differentiable with respect to learned latent parameters. We prove that the associated optimizer map is single-valued, maps onto the relative interior of the feasible set, admits an explicit Jacobian, is Lipschitz continuous, and can be made arbitrarily smooth. We also establish a universal approximation result showing that the proposed class can approximate any continuous feasible policy on compact context sets. The framework unifies explicitly regularized optimizers and implicit perturbation-based smooth optimizers. Experiments on contextual newsvendor and resource allocation problems show that our approach improves prescriptive performance relative to the benchmark methods.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

92 extracted references · 1 canonical work pages

  1. [1]

    predict, then optimize

    Smart “predict, then optimize” , author=. Management Science , volume=. 2022 , publisher=

    2022

  2. [2]

    Management Science , volume=

    Fast rates for contextual linear optimization , author=. Management Science , volume=. 2022 , publisher=

    2022

  3. [3]

    2013 , publisher=

    Convex analysis and minimization algorithms I: Fundamentals , author=. 2013 , publisher=

    2013

  4. [4]

    Neural networks , volume=

    On the approximate realization of continuous mappings by neural networks , author=. Neural networks , volume=. 1989 , publisher=

    1989

  5. [5]

    Neural networks , volume=

    Multilayer feedforward networks with a nonpolynomial activation function can approximate any function , author=. Neural networks , volume=. 1993 , publisher=

    1993

  6. [6]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Melding the data-decisions pipeline: Decision-focused learning for combinatorial optimization , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  7. [7]

    Mathematical Programming Computation , volume=

    PyEPO: a PyTorch-based end-to-end predict-then-optimize library for linear and integer programming , author=. Mathematical Programming Computation , volume=. 2024 , publisher=

    2024

  8. [8]

    Advances in Neural Information Processing Systems , volume=

    Decision-focused learning with directional gradients , author=. Advances in Neural Information Processing Systems , volume=

Show all 92 references
  1. [9]

    Niculae, Vlad and Martins, Andre and Blondel, Mathieu and Cardie, Claire , booktitle=. Sparse. 2018 , organization=

    2018

  2. [10]

    Mathematical programming , volume=

    Smooth minimization of non-smooth functions , author=. Mathematical programming , volume=. 2005 , publisher=

    2005

  3. [11]

    Conference on Learning Theory , pages=

    Online linear optimization via smoothing , author=. Conference on Learning Theory , pages=. 2014 , organization=

    2014

  4. [12]

    Advances in Neural Information Processing Systems , volume=

    Learning with differentiable perturbed optimizers , author=. Advances in Neural Information Processing Systems , volume=

  5. [13]

    1970 , publisher=

    Convex Analysis , author=. 1970 , publisher=

    1970

  6. [14]

    European Journal of Operational Research , volume=

    A survey of contextual optimization methods for decision-making under uncertainty , author=. European Journal of Operational Research , volume=. 2025 , publisher=

    2025

  7. [15]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages=

    Deep network flow for multi-object tracking , author=. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages=

  8. [16]

    arXiv preprint arXiv:2111.09170 , year=

    A universal end-to-end approach to portfolio optimization via deep learning , author=. arXiv preprint arXiv:2111.09170 , year=

    arXiv

  9. [17]

    IEEE Transactions on Power Systems , volume=

    End-to-end feasible optimization proxies for large-scale economic dispatch , author=. IEEE Transactions on Power Systems , volume=. 2023 , publisher=

    2023

  10. [18]

    Operations Research , volume=

    The big data newsvendor: Practical insights from machine learning , author=. Operations Research , volume=. 2019 , publisher=

    2019

  11. [19]

    Management Science , volume=

    From predictive to prescriptive analytics , author=. Management Science , volume=. 2020 , publisher=

    2020

  12. [20]

    On Data-Driven Prescriptive Analytics with Side Information: A Regularized

    Wang, Yijie and Srivastava, Prateek R and Hanasusanto, Grani A and Ho, Chin Pang , journal=. On Data-Driven Prescriptive Analytics with Side Information: A Regularized. 2026 , publisher=

    2026

  13. [21]

    Operations Research , volume=

    Data-driven sample average approximation with covariate information , author=. Operations Research , volume=. 2025 , publisher=

    2025

  14. [22]

    Mathematical Programming , volume=

    Residuals-based distributionally robust optimization with covariate information , author=. Mathematical Programming , volume=. 2024 , publisher=

    2024

  15. [23]

    2022 , eprint=

    Data-Driven Sample Average Approximation with Covariate Information , author=. 2022 , eprint=

    2022

  16. [24]

    Advances in Neural Information Processing Systems , volume=

    Differentiable convex optimization layers , author=. Advances in Neural Information Processing Systems , volume=

  17. [25]

    Advances in Neural Information Processing Systems , volume=

    Task-based end-to-end model learning in stochastic optimization , author=. Advances in Neural Information Processing Systems , volume=

  18. [26]

    International Conference on Machine Learning , pages=

    Optnet: Differentiable optimization as a layer in neural networks , author=. International Conference on Machine Learning , pages=. 2017 , organization=

    2017

  19. [27]

    Operations Research , volume=

    Data-driven optimization: A reproducing kernel hilbert space approach , author=. Operations Research , volume=. 2022 , publisher=

    2022

  20. [28]

    Operations Research , year=

    Data-driven piecewise affine decision rules for stochastic programming with covariate information , author=. Operations Research , year=

  21. [29]

    Management Science , volume=

    A practical end-to-end inventory management model with deep learning , author=. Management Science , volume=. 2023 , publisher=

    2023

  22. [31]

    Iise Transactions , volume=

    Applying deep learning to the newsvendor problem , author=. Iise Transactions , volume=. 2020 , publisher=

    2020

  23. [32]

    Advances in Neural Information Processing Systems , volume=

    Nonparametric density estimation for stochastic optimization with an observable state variable , author=. Advances in Neural Information Processing Systems , volume=

  24. [33]

    European Journal of Operational Research , volume=

    Dynamic optimization with side information , author=. European Journal of Operational Research , volume=. 2023 , publisher=

    2023

  25. [34]

    The Review of Financial Studies , volume=

    Parametric portfolio policies: Exploiting characteristics in the cross-section of equity returns , author=. The Review of Financial Studies , volume=. 2009 , publisher=

    2009

  26. [35]

    INFOR: Information Systems and Operational Research , volume=

    Generalization bounds for regularized portfolio selection with market side information , author=. INFOR: Information Systems and Operational Research , volume=. 2020 , publisher=

    2020

  27. [36]

    arXiv preprint arXiv:2509.14557 , year=

    Data-Driven Contextual Optimization with Gaussian Mixtures: Flow-Based Generalization, Robust Models, and Multistage Extensions , author=. arXiv preprint arXiv:2509.14557 , year=

  28. [37]

    Operations Research , year=

    Integrated conditional estimation-optimization , author=. Operations Research , year=

  29. [38]

    Advances in Neural Information Processing Systems , volume=

    Nonsmooth implicit differentiation for machine-learning and optimization , author=. Advances in Neural Information Processing Systems , volume=

  30. [39]

    Available at SSRN 3623006 , year=

    Decision-driven regularization: A blended model for predict-then-optimize , author=. Available at SSRN 3623006 , year=

  31. [40]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Smart predict-and-optimize for hard combinatorial optimization problems , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  32. [41]

    International Conference on Machine Learning , pages=

    Decision trees for decision-making under the predict-then-optimize framework , author=. International Conference on Machine Learning , pages=. 2020 , organization=

    2020

  33. [42]

    Management Science , volume=

    Stochastic Optimization Forests , author=. Management Science , volume=. 2023 , publisher=

    2023

  34. [43]

    Unpublished Manuscript, http://ttic

    On the duality of strong convexity and strong smoothness: Learning applications and matrix regularization , author=. Unpublished Manuscript, http://ttic. uchicago. edu/shai/papers/KakadeShalevTewari09. pdf , volume=

  35. [44]

    2009 , publisher=

    Variational Analysis , author=. 2009 , publisher=

    2009

  36. [46]

    Practical

    Snoek, Jasper and Larochelle, Hugo and Adams, Ryan P , journal=. Practical

  37. [47]

    Journal of Global optimization , volume=

    Efficient global optimization of expensive black-box functions , author=. Journal of Global optimization , volume=. 1998 , publisher=

    1998

  38. [48]

    International Conference on Artificial Intelligence and Statistics , pages=

    Mind the (optimality) gap: a gap-aware learning rate scheduler for adversarial nets , author=. International Conference on Artificial Intelligence and Statistics , pages=. 2023 , organization=

    2023

  39. [49]

    Advances in Neural Information Processing Systems , volume=

    Calibration by distribution matching: Trainable kernel calibration metrics , author=. Advances in Neural Information Processing Systems , volume=

  40. [50]

    Advances in Neural Information Processing Systems , volume=

    Risk bounds and calibration for a smart predict-then-optimize method , author=. Advances in Neural Information Processing Systems , volume=

  41. [51]

    , title =

    Bertsekas, Dimitri P. , title =

  42. [52]

    , author=

    Control of uncertain systems with a set-membership description of the uncertainty. , author=. 1971 , school=

    1971

  43. [53]

    Learning with

    Blondel, Mathieu and Martins, Andr. Learning with. Journal of Machine Learning Research , volume=

  44. [54]

    Machine learning meets

    Wang, Yijie and Gao, Hao and Harvey, Campbell R and Liu, Yan and Tao, Xinyuan , year=. Machine learning meets

  45. [55]

    Online linear optimization via smoothing

    Jacob Abernethy, Chansoo Lee, Abhinav Sinha, and Ambuj Tewari. Online linear optimization via smoothing. In Conference on Learning Theory, pages 807--823. PMLR, 2014

    2014

  46. [56]

    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 Systems, 32, 2019

    2019

  47. [57]

    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

  48. [58]

    The big data newsvendor: Practical insights from machine learning

    Gah-Yi Ban and Cynthia Rudin. The big data newsvendor: Practical insights from machine learning. Operations Research, 67 0 (1): 0 90--108, 2019

    2019

  49. [59]

    Generalization bounds for regularized portfolio selection with market side information

    Thierry Bazier-Matte and Erick Delage. Generalization bounds for regularized portfolio selection with market side information. INFOR: Information Systems and Operational Research, 58 0 (2): 0 374--401, 2020

    2020

  50. [60]

    Learning with differentiable perturbed optimizers

    Quentin Berthet, Mathieu Blondel, Olivier Teboul, Marco Cuturi, Jean-Philippe Vert, and Francis Bach. Learning with differentiable perturbed optimizers. Advances in Neural Information Processing Systems, 33: 0 9508--9519, 2020

    2020

  51. [61]

    Control of uncertain systems with a set-membership description of the uncertainty

    Dimitri P Bertsekas. Control of uncertain systems with a set-membership description of the uncertainty. PhD thesis, Massachusetts Institute of Technology, 1971

    1971

  52. [62]

    From predictive to prescriptive analytics

    Dimitris Bertsimas and Nathan Kallus. From predictive to prescriptive analytics. Management Science, 66 0 (3): 0 1025--1044, 2020

    2020

  53. [63]

    Data-driven optimization: A reproducing kernel hilbert space approach

    Dimitris Bertsimas and Nihal Koduri. Data-driven optimization: A reproducing kernel hilbert space approach. Operations Research, 70 0 (1): 0 454--471, 2022

    2022

  54. [64]

    Dynamic optimization with side information

    Dimitris Bertsimas, Christopher McCord, and Bradley Sturt. Dynamic optimization with side information. European Journal of Operational Research, 304 0 (2): 0 634--651, 2023

    2023

  55. [65]

    Learning with F enchel- Y oung losses

    Mathieu Blondel, Andr \'e FT Martins, and Vlad Niculae. Learning with F enchel- Y oung losses. Journal of Machine Learning Research, 21 0 (35): 0 1--69, 2020

    2020

  56. [66]

    Nonsmooth implicit differentiation for machine-learning and optimization

    J \'e r \^o me Bolte, Tam Le, Edouard Pauwels, and Tony Silveti-Falls. Nonsmooth implicit differentiation for machine-learning and optimization. Advances in Neural Information Processing Systems, 34: 0 13537--13549, 2021

    2021

  57. [67]

    Parametric portfolio policies: Exploiting characteristics in the cross-section of equity returns

    Michael W Brandt, Pedro Santa-Clara, and Rossen Valkanov. Parametric portfolio policies: Exploiting characteristics in the cross-section of equity returns. The Review of Financial Studies, 22 0 (9): 0 3411--3447, 2009

    2009

  58. [68]

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

    Wenbo Chen, Mathieu Tanneau, and Pascal Van Hentenryck. End-to-end feasible optimization proxies for large-scale economic dispatch. IEEE Transactions on Power Systems, 39 0 (2): 0 4723--4734, 2023

    2023

  59. [69]

    Task-based end-to-end model learning in stochastic optimization

    Priya Donti, Brandon Amos, and J Zico Kolter. Task-based end-to-end model learning in stochastic optimization. Advances in Neural Information Processing Systems, 30, 2017

    2017

  60. [70]

    predict, then optimize

    Adam N Elmachtoub and Paul Grigas. Smart “predict, then optimize”. Management Science, 68 0 (1): 0 9--26, 2022

    2022

  61. [71]

    Decision trees for decision-making under the predict-then-optimize framework

    Adam N Elmachtoub, Jason Cheuk Nam Liang, and Ryan McNellis. Decision trees for decision-making under the predict-then-optimize framework. In International Conference on Machine Learning, pages 2858--2867. PMLR, 2020

    2020

  62. [72]

    Mind the (optimality) gap: a gap-aware learning rate scheduler for adversarial nets

    Hussein Hazimeh and Natalia Ponomareva. Mind the (optimality) gap: a gap-aware learning rate scheduler for adversarial nets. In International Conference on Artificial Intelligence and Statistics, pages 3018--3033. PMLR, 2023

    2023

  63. [73]

    Efficient global optimization of expensive black-box functions

    Donald R Jones, Matthias Schonlau, and William J Welch. Efficient global optimization of expensive black-box functions. Journal of Global optimization, 13 0 (4): 0 455--492, 1998

    1998

  64. [74]

    Stochastic optimization forests

    Nathan Kallus and Xiaojie Mao. Stochastic optimization forests. Management Science, 69 0 (4): 0 1975--1994, 2023

    1975

  65. [75]

    Rohit Kannan, G \"u zin Bayraksan, and James R. Luedtke. Data-driven sample average approximation with covariate information, 2022

    2022

  66. [76]

    Rohit Kannan, G \"u zin Bayraksan, and James R. Luedtke. Residuals-based distributionally robust optimization with covariate information. Mathematical Programming, 207 0 (1): 0 369--425, 2024. doi:10.1007/s10107-023-02016-z

    2024 doi

  67. [77]

    Data-driven sample average approximation with covariate information

    Rohit Kannan, G \"u zin Bayraksan, and James R Luedtke. Data-driven sample average approximation with covariate information. Operations Research, 73 0 (6): 0 3245--3259, 2025

    2025

  68. [78]

    Risk bounds and calibration for a smart predict-then-optimize method

    Heyuan Liu and Paul Grigas. Risk bounds and calibration for a smart predict-then-optimize method. In Advances in Neural Information Processing Systems, volume 34, pages 22083--22094, 2021

    2021

  69. [79]

    Decision-driven regularization: A blended model for predict-then-optimize

    Gar Goei Loke, Qinshen Tang, and Yangge Xiao. Decision-driven regularization: A blended model for predict-then-optimize. Available at SSRN 3623006, 2022

    2022

  70. [80]

    Smart predict-and-optimize for hard combinatorial optimization problems

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

    2020

  71. [81]

    Calibration by distribution matching: Trainable kernel calibration metrics

    Charlie Marx, Sofian Zalouk, and Stefano Ermon. Calibration by distribution matching: Trainable kernel calibration metrics. Advances in Neural Information Processing Systems, 36: 0 25910--25928, 2023

    2023

  72. [82]

    Smooth minimization of non-smooth functions

    Yurii Nesterov. Smooth minimization of non-smooth functions. Mathematical programming, 103 0 (1): 0 127--152, 2005

    2005

  73. [83]

    Sparse MAP : Differentiable sparse structured inference

    Vlad Niculae, Andre Martins, Mathieu Blondel, and Claire Cardie. Sparse MAP : Differentiable sparse structured inference. In International Conference on Machine Learning, pages 3799--3808. PMLR, 2018

    2018

  74. [84]

    Applying deep learning to the newsvendor problem

    Afshin Oroojlooyjadid, Lawrence V Snyder, and Martin Tak \'a c . Applying deep learning to the newsvendor problem. Iise Transactions, 52 0 (4): 0 444--463, 2020

    2020

  75. [85]

    A practical end-to-end inventory management model with deep learning

    Meng Qi, Yuanyuan Shi, Yongzhi Qi, Chenxin Ma, Rong Yuan, Di Wu, and Zuo-Jun Shen. A practical end-to-end inventory management model with deep learning. Management Science, 69 0 (2): 0 759--773, 2023

    2023

  76. [86]

    Tyrrell Rockafellar

    R. Tyrrell Rockafellar. Convex Analysis, volume 18. Princeton University Press, 1970

    1970

  77. [87]

    Tyrrell Rockafellar and Roger J.-B

    R. Tyrrell Rockafellar and Roger J.-B. Wets. Variational Analysis, volume 317. Springer Science & Business Media, Berlin, Heidelberg, 2009. ISBN 978-3-540-62772-2

    2009

  78. [88]

    Schneider and Daniel Kuhn

    Philipp J. Schneider and Daniel Kuhn. Soft-radial projection for constrained end-to-end learning. arXiv preprint arXiv:2602.03461, 2026

    2026

  79. [89]

    Practical B ayesian optimization of machine learning algorithms

    Jasper Snoek, Hugo Larochelle, and Ryan P Adams. Practical B ayesian optimization of machine learning algorithms. Advances in Neural Information Processing Systems, 25, 2012

    2012

  80. [90]

    Machine learning meets M arkowitz

    Yijie Wang, Hao Gao, Campbell R Harvey, Yan Liu, and Xinyuan Tao. Machine learning meets M arkowitz. Technical report, National Bureau of Economic Research, 2026 a

    2026

  81. [91]

    On data-driven prescriptive analytics with side information: A regularized N adaraya-- W atson approach

    Yijie Wang, Prateek R Srivastava, Grani A Hanasusanto, and Chin Pang Ho. On data-driven prescriptive analytics with side information: A regularized N adaraya-- W atson approach. Manufacturing & Service Operations Management, 2026 b

    2026

  82. [92]

    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

    2019

  83. [93]

    Data-driven piecewise affine decision rules for stochastic programming with covariate information

    Yiyang Zhang, Junyi Liu, and Xiaobo Zhao. Data-driven piecewise affine decision rules for stochastic programming with covariate information. Operations Research, 2026

    2026

  84. [94]

    Deep learning for portfolio optimization

    Zihao Zhang, Stefan Zohren, and Stephen Roberts. Deep learning for portfolio optimization. arXiv preprint arXiv:2005.13665, 2020

    2005 arXiv

Pith tools

Reviewed July 31, 2026 · model on record in the stance chip above.