Wasserstein Distributionally Robust Risk-Sensitive Estimation via Conditional Value-at-Risk

Eilyan Bitar; Feras Al Taha

arxiv: 2604.18546 · v1 · submitted 2026-04-20 · 💻 cs.LG · eess.SP· math.OC

Wasserstein Distributionally Robust Risk-Sensitive Estimation via Conditional Value-at-Risk

Feras Al Taha , Eilyan Bitar This is my paper

Pith reviewed 2026-05-10 05:18 UTC · model grok-4.3

classification 💻 cs.LG eess.SPmath.OC

keywords Wasserstein distancedistributionally robust optimizationconditional value-at-risksemidefinite programmingaffine estimationrisk-sensitive estimationelectricity price forecasting

0 comments

The pith

Affine estimators minimizing worst-case CVaR of squared error over a Wasserstein ball are exactly computable by semidefinite programming when the nominal distribution is finitely supported.

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

The paper proposes a distributionally robust method for estimating an unknown signal from an observation when their joint distribution is uncertain but lies inside a type-2 Wasserstein ball around a given nominal distribution. Performance is measured by the conditional value-at-risk of the squared estimation error, and the goal is to find affine estimators that minimize this risk in the worst case over the ball. The central result is that this min-max problem reduces exactly to a tractable semidefinite program whenever the nominal distribution has finite support. The resulting estimators are tested on real wholesale electricity price data and shown to achieve lower out-of-sample CVaR than standard approaches.

Core claim

When the nominal distribution at the center of the Wasserstein ball is finitely supported, the affine estimator that minimizes the worst-case conditional value-at-risk of squared error over all distributions inside the ball can be recovered exactly by solving a semidefinite program.

What carries the argument

The exact reduction of the distributionally robust CVaR minimization problem over Wasserstein ambiguity sets to a semidefinite program, which holds under finite support of the nominal distribution.

If this is right

The estimators provide explicit performance guarantees against any distribution shift inside the chosen Wasserstein ball.
Computation becomes practical for any finite-support nominal using off-the-shelf SDP solvers.
On forecasting tasks the method yields lower realized CVaR of squared error than non-robust or alternative robust estimators.
The framework applies directly to any linear estimation setting where squared-error risk is evaluated via CVaR.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same SDP reduction technique may apply to other convex risk measures or to non-affine estimator classes under similar ambiguity sets.
Finite-support assumptions could be relaxed in practice by discretizing continuous nominal distributions and controlling the resulting approximation error.
The approach suggests a general template for turning Wasserstein-robust risk-sensitive problems into convex programs in signal processing and time-series prediction.

Load-bearing premise

The nominal distribution at the center of the Wasserstein ball must have finite support for the semidefinite program to recover the exact optimal estimator.

What would settle it

A concrete counterexample with an infinitely supported nominal distribution in which the semidefinite program solution differs from the true optimal estimator, or an empirical test on forecasting data where the method fails to produce lower out-of-sample CVaR than non-robust baselines.

Figures

Figures reproduced from arXiv: 2604.18546 by Eilyan Bitar, Feras Al Taha.

read the original abstract

We propose a distributionally robust approach to risk-sensitive estimation of an unknown signal x from an observed signal y. The unknown signal and observation are modeled as random vectors whose joint probability distribution is unknown, but assumed to belong to a given type-2 Wasserstein ball of distributions, termed the ambiguity set. The performance of an estimator is measured according to the conditional value-at-risk (CVaR) of the squared estimation error. Within this framework, we study the problem of computing affine estimators that minimize the worst-case CVaR over all distributions in the given ambiguity set. As our main result, we show that, when the nominal distribution at the center of the Wasserstein ball is finitely supported, such estimators can be exactly computed by solving a tractable semidefinite program. We evaluate the proposed estimators on a wholesale electricity price forecasting task using real market data and show that they deliver lower out-of-sample CVaR of squared error compared to existing methods.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This paper reduces worst-case CVaR affine estimation over a Wasserstein ball to a tractable SDP when the nominal distribution has finite support, and tests the result on electricity price data.

read the letter

The core contribution is a reduction showing that the min-max problem for affine estimators minimizing CVaR of squared error, with the worst case taken over a type-2 Wasserstein ball, becomes an SDP when the center distribution is finitely supported. That is the main new piece: a concrete computational handle on the combination of distributional robustness and CVaR risk sensitivity for this estimation setting. The electricity forecasting experiment then applies the estimator to real market data and reports lower out-of-sample CVaR than some existing methods, which at least demonstrates that the formulation can be solved and used on a practical task. The finite-support condition is stated up front, so the claim stays within its stated scope rather than overreaching. The modeling choice that the true distribution lies inside the ball is standard for this line of work and does not create an internal contradiction with the SDP result. On the downside, the finite-support restriction means the exact reduction does not apply directly to continuous nominal distributions, which will often require discretization or other approximations in practice. The numerical section is limited to one forecasting task, so it is not yet clear how the method scales or performs across a wider range of estimation problems. The paper stays within the DRO and risk-measure literature without claiming broader impact outside that area. Readers working on robust optimization or risk-sensitive signal processing would find the SDP formulation useful as a building block. It is worth sending to peer review because the central reduction is a clean, verifiable computational result paired with a real-data check, even if the experiments remain narrow.

Referee Report

0 major / 3 minor

Summary. The paper proposes a distributionally robust framework for risk-sensitive estimation of an unknown signal x from an observation y. The joint distribution is assumed to lie in a type-2 Wasserstein ball centered at a nominal distribution. Estimator performance is measured by the conditional value-at-risk (CVaR) of the squared error, and the focus is on finding affine estimators that minimize the worst-case CVaR over the ambiguity set. The central claim is that when the nominal distribution has finite support, the resulting min-max problem reduces exactly to a tractable semidefinite program (SDP). The approach is evaluated on a wholesale electricity price forecasting task with real market data, where the proposed estimators achieve lower out-of-sample CVaR compared to existing methods.

Significance. If the SDP reduction holds, the work provides a computationally tractable solution for distributionally robust, risk-sensitive estimation under Wasserstein ambiguity, which is valuable for applications involving uncertainty quantification such as energy forecasting. The explicit finite-support condition enabling the exact SDP reformulation is a clear strength, as is the use of real-world data for validation. This bridges tools from distributionally robust optimization and CVaR in a way that yields practical estimators without requiring full knowledge of the true distribution.

minor comments (3)

The abstract and introduction could more explicitly state the dimensions of the SDP (e.g., number of variables and constraints in terms of support size) to better convey computational scalability.
In the numerical experiments, include a sensitivity analysis with respect to the Wasserstein radius to demonstrate robustness of the performance gains.
Clarify the choice of affine estimator class versus more general estimators and whether the SDP result extends beyond affine forms.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment of our manuscript, the recognition of its contributions to distributionally robust risk-sensitive estimation, and the recommendation for minor revision. No major comments were provided in the report.

Circularity Check

0 steps flagged

No significant circularity detected in derivation chain

full rationale

The paper's central claim is a conditional mathematical reduction: under the explicit assumption that the nominal distribution is finitely supported, the affine estimator minimizing worst-case CVaR over the Wasserstein ball is exactly equivalent to a tractable SDP. This is presented as a derived computational result (not a redefinition or tautology). No self-definitional loops appear, no fitted parameters are renamed as predictions, and no load-bearing self-citations or uniqueness theorems imported from the authors' prior work are invoked in the abstract or main result. The finite-support condition is stated as necessary for the exact SDP equivalence, and the modeling assumption that the true distribution lies in the ball is standard for DRO without creating internal inconsistency. The derivation chain is self-contained against external benchmarks and does not reduce to its inputs by construction.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The framework rests on the standard assumption that the unknown joint distribution belongs to a Wasserstein ball of specified radius around a nominal distribution; the radius itself is a free parameter that must be chosen externally.

free parameters (1)

Wasserstein ball radius
The radius controls the size of the ambiguity set and must be selected by the user, typically via validation or domain knowledge; it is not derived from the data within the paper.

axioms (1)

domain assumption The true joint distribution of the signal and observation lies inside the given type-2 Wasserstein ball centered at the nominal distribution.
This defines the ambiguity set and is invoked throughout the distributionally robust formulation.

pith-pipeline@v0.9.0 · 5471 in / 1403 out tokens · 35004 ms · 2026-05-10T05:18:41.348505+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages

[1]

A minimax Chebyshev estimator for bounded error estimation,

Y . C. Eldar, A. Beck, and M. Teboulle, “A minimax Chebyshev estimator for bounded error estimation,”IEEE Transactions on Signal Processing, vol. 56, no. 4, pp. 1388–1397, 2008

work page 2008
[2]

Near-optimality of linear recovery in Gaussian observation scheme under∥ · ∥ 2 2-loss,

A. Juditsky and A. Nemirovski, “Near-optimality of linear recovery in Gaussian observation scheme under∥ · ∥ 2 2-loss,”The Annals of Statistics, vol. 46, no. 4, pp. 1603–1629, 2018

work page 2018
[3]

Regularization in regression with bounded noise: A Chebyshev center approach,

A. Beck and Y . C. Eldar, “Regularization in regression with bounded noise: A Chebyshev center approach,”SIAM Journal on Matrix Analysis and Applications, vol. 29, no. 2, pp. 606–625, 2007

work page 2007
[4]

Robust mean-squared error estimation in the presence of model uncertainties,

Y . C. Eldar, A. Ben-Tal, and A. Nemirovski, “Robust mean-squared error estimation in the presence of model uncertainties,”IEEE Trans- actions on Signal Processing, vol. 53, no. 1, pp. 168–181, 2004

work page 2004
[5]

Mean-squared error estimation for linear systems with block circulant uncertainty,

A. Beck, Y . C. Eldar, and A. Ben-Tal, “Mean-squared error estimation for linear systems with block circulant uncertainty,”SIAM Journal on Matrix Analysis and Applications, vol. 29, no. 3, pp. 712–730, 2007

work page 2007
[6]

Robust competitive estimation with signal and noise covariance uncertainties,

Y . C. Eldar, “Robust competitive estimation with signal and noise covariance uncertainties,”IEEE Transactions on Information Theory, vol. 52, no. 10, pp. 4532–4547, 2006

work page 2006
[7]

Robust mean-squared error estimation of multiple signals in linear systems affected by model and noise uncertainties,

A. Beck, A. Ben-Tal, and Y . C. Eldar, “Robust mean-squared error estimation of multiple signals in linear systems affected by model and noise uncertainties,”Mathematical Programming, vol. 107, no. 1, pp. 155–187, 2006

work page 2006
[8]

Robust least-squares estimation with a relative entropy constraint,

B. C. Levy and R. Nikoukhah, “Robust least-squares estimation with a relative entropy constraint,”IEEE Transactions on Information Theory, vol. 50, no. 1, pp. 89–104, 2004

work page 2004
[9]

On the robustness of the Bayes and Wiener estimators under model uncertainty,

M. Zorzi, “On the robustness of the Bayes and Wiener estimators under model uncertainty,”Automatica, vol. 83, pp. 133–140, 2017

work page 2017
[10]

Bridging Bayesian and minimax mean square error estimation via Wasserstein distributionally robust optimization,

V . A. Nguyen, S. Shafieezadeh-Abadeh, D. Kuhn, and P. Mohajerin Es- fahani, “Bridging Bayesian and minimax mean square error estimation via Wasserstein distributionally robust optimization,”Mathematics of Operations Research, vol. 48, no. 1, pp. 1–37, 2023

work page 2023
[11]

Wasserstein distributionally robust Kalman filtering,

S. Shafieezadeh Abadeh, V . A. Nguyen, D. Kuhn, and P. M. Moha- jerin Esfahani, “Wasserstein distributionally robust Kalman filtering,” Advances in Neural Information Processing Systems, vol. 31, 2018

work page 2018
[12]

Distributionally robust logistic regression,

S. Shafieezadeh Abadeh, P. M. Mohajerin Esfahani, and D. Kuhn, “Distributionally robust logistic regression,”Advances in Neural In- formation Processing Systems, vol. 28, 2015

work page 2015
[13]

Regularization via mass transportation,

S. Shafieezadeh-Abadeh, D. Kuhn, and P. M. Esfahani, “Regularization via mass transportation,”Journal of Machine Learning Research, vol. 20, no. 103, pp. 1–68, 2019

work page 2019
[14]

Wasserstein distributionally robust estimation in high dimensions: performance analysis and opti- mal hyperparameter tuning,

L. Aolaritei, S. Shafiee, and F. D ¨orfler, “Wasserstein distributionally robust estimation in high dimensions: performance analysis and opti- mal hyperparameter tuning,”Math. Programming, pp. 1–85, 2026

work page 2026
[15]

Sequential domain adaptation by synthesizing distributionally robust experts,

B. Taskesen, M.-C. Yue, J. Blanchet, D. Kuhn, and V . A. Nguyen, “Sequential domain adaptation by synthesizing distributionally robust experts,” inInternational Conference on Machine Learning. PMLR, 2021, pp. 10 162–10 172

work page 2021
[16]

Risk-aversion in multi-armed bandits,

A. Sani, A. Lazaric, and R. Munos, “Risk-aversion in multi-armed bandits,”Advances in Neural Info. Processing Systems, vol. 25, 2012

work page 2012
[17]

Risk-sensitive and robust decision-making: a CVaR optimization approach,

Y . Chow, A. Tamar, S. Mannor, and M. Pavone, “Risk-sensitive and robust decision-making: a CVaR optimization approach,”Advances in Neural Information Processing Systems, vol. 28, 2015

work page 2015
[18]

Risk- constrained reinforcement learning with percentile risk criteria,

Y . Chow, M. Ghavamzadeh, L. Janson, and M. Pavone, “Risk- constrained reinforcement learning with percentile risk criteria,”Jour- nal of Machine Learning Research, vol. 18, no. 167, pp. 1–51, 2018

work page 2018
[19]

Fairness risk measures,

R. Williamson and A. Menon, “Fairness risk measures,” inInterna- tional Conf. on Machine Learning. PMLR, 2019, pp. 6786–6797

work page 2019
[20]

Statistical learning with conditional value at risk,

T. Soma and Y . Yoshida, “Statistical learning with conditional value at risk,”arXiv preprint arXiv:2002.05826, 2020

work page arXiv 2002
[21]

Superquantiles at work: Machine learning applications and efficient subgradient computation,

Y . Laguel, K. Pillutla, J. Malick, and Z. Harchaoui, “Superquantiles at work: Machine learning applications and efficient subgradient computation,”Set-Valued and Variational Analysis, vol. 29, no. 4, pp. 967–996, 2021

work page 2021
[22]

Distri- butionally robust control of constrained stochastic systems,

B. P. Van Parys, D. Kuhn, P. J. Goulart, and M. Morari, “Distri- butionally robust control of constrained stochastic systems,”IEEE Transactions on Automatic Control, vol. 61, no. 2, pp. 430–442, 2015

work page 2015
[23]

Toward a scalable upper bound for a CVaR-LQ problem,

M. P. Chapman and L. Lessard, “Toward a scalable upper bound for a CVaR-LQ problem,”IEEE Control Systems Letters, vol. 6, pp. 920– 925, 2021

work page 2021
[24]

Risk-aware linear quadratic control using conditional value-at-risk,

M. Kishida and A. Cetinkaya, “Risk-aware linear quadratic control using conditional value-at-risk,”IEEE Transactions on Automatic Control, vol. 68, no. 1, pp. 416–423, 2022

work page 2022
[25]

arXiv preprint arXiv:2112.09959 , title =

V . A. Nguyen, S. Shafiee, D. Filipovi ´c, and D. Kuhn, “Mean-covariance robust risk measurement,”arXiv preprint arXiv:2112.09959, 2021

work page arXiv 2021
[26]

Distributionally robust optimization,

D. Kuhn, S. Shafiee, and W. Wiesemann, “Distributionally robust optimization,”Acta Numerica, vol. 34, p. 579–804, 2025

work page 2025
[27]

Wasserstein distributionally robust optimization: Theory and appli- cations in machine learning,

D. Kuhn, P. M. Esfahani, V . A. Nguyen, and S. Shafieezadeh-Abadeh, “Wasserstein distributionally robust optimization: Theory and appli- cations in machine learning,” inOperations research & management science in the age of analytics. Informs, 2019, pp. 130–166

work page 2019
[28]

A distributionally robust approach to regret optimal control using the Wasserstein distance,

F. Al Taha, S. Yan, and E. Bitar, “A distributionally robust approach to regret optimal control using the Wasserstein distance,” in62nd IEEE Conference on Decision and Control. IEEE, 2023, pp. 2768–2775

work page 2023
[29]

On a formula for theL 2 Wasserstein metric be- tween measures on Euclidean and Hilbert spaces,

M. Gelbrich, “On a formula for theL 2 Wasserstein metric be- tween measures on Euclidean and Hilbert spaces,”Mathematische Nachrichten, vol. 147, no. 1, pp. 185–203, 1990

work page 1990
[30]

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

P. Donti, B. Amos, and J. Z. Kolter, “Task-based end-to-end model learning in stochastic optimization,”Advances in Neural Information Processing Systems, vol. 30, 2017

work page 2017

[1] [1]

A minimax Chebyshev estimator for bounded error estimation,

Y . C. Eldar, A. Beck, and M. Teboulle, “A minimax Chebyshev estimator for bounded error estimation,”IEEE Transactions on Signal Processing, vol. 56, no. 4, pp. 1388–1397, 2008

work page 2008

[2] [2]

Near-optimality of linear recovery in Gaussian observation scheme under∥ · ∥ 2 2-loss,

A. Juditsky and A. Nemirovski, “Near-optimality of linear recovery in Gaussian observation scheme under∥ · ∥ 2 2-loss,”The Annals of Statistics, vol. 46, no. 4, pp. 1603–1629, 2018

work page 2018

[3] [3]

Regularization in regression with bounded noise: A Chebyshev center approach,

A. Beck and Y . C. Eldar, “Regularization in regression with bounded noise: A Chebyshev center approach,”SIAM Journal on Matrix Analysis and Applications, vol. 29, no. 2, pp. 606–625, 2007

work page 2007

[4] [4]

Robust mean-squared error estimation in the presence of model uncertainties,

Y . C. Eldar, A. Ben-Tal, and A. Nemirovski, “Robust mean-squared error estimation in the presence of model uncertainties,”IEEE Trans- actions on Signal Processing, vol. 53, no. 1, pp. 168–181, 2004

work page 2004

[5] [5]

Mean-squared error estimation for linear systems with block circulant uncertainty,

A. Beck, Y . C. Eldar, and A. Ben-Tal, “Mean-squared error estimation for linear systems with block circulant uncertainty,”SIAM Journal on Matrix Analysis and Applications, vol. 29, no. 3, pp. 712–730, 2007

work page 2007

[6] [6]

Robust competitive estimation with signal and noise covariance uncertainties,

Y . C. Eldar, “Robust competitive estimation with signal and noise covariance uncertainties,”IEEE Transactions on Information Theory, vol. 52, no. 10, pp. 4532–4547, 2006

work page 2006

[7] [7]

Robust mean-squared error estimation of multiple signals in linear systems affected by model and noise uncertainties,

A. Beck, A. Ben-Tal, and Y . C. Eldar, “Robust mean-squared error estimation of multiple signals in linear systems affected by model and noise uncertainties,”Mathematical Programming, vol. 107, no. 1, pp. 155–187, 2006

work page 2006

[8] [8]

Robust least-squares estimation with a relative entropy constraint,

B. C. Levy and R. Nikoukhah, “Robust least-squares estimation with a relative entropy constraint,”IEEE Transactions on Information Theory, vol. 50, no. 1, pp. 89–104, 2004

work page 2004

[9] [9]

On the robustness of the Bayes and Wiener estimators under model uncertainty,

M. Zorzi, “On the robustness of the Bayes and Wiener estimators under model uncertainty,”Automatica, vol. 83, pp. 133–140, 2017

work page 2017

[10] [10]

Bridging Bayesian and minimax mean square error estimation via Wasserstein distributionally robust optimization,

V . A. Nguyen, S. Shafieezadeh-Abadeh, D. Kuhn, and P. Mohajerin Es- fahani, “Bridging Bayesian and minimax mean square error estimation via Wasserstein distributionally robust optimization,”Mathematics of Operations Research, vol. 48, no. 1, pp. 1–37, 2023

work page 2023

[11] [11]

Wasserstein distributionally robust Kalman filtering,

S. Shafieezadeh Abadeh, V . A. Nguyen, D. Kuhn, and P. M. Moha- jerin Esfahani, “Wasserstein distributionally robust Kalman filtering,” Advances in Neural Information Processing Systems, vol. 31, 2018

work page 2018

[12] [12]

Distributionally robust logistic regression,

S. Shafieezadeh Abadeh, P. M. Mohajerin Esfahani, and D. Kuhn, “Distributionally robust logistic regression,”Advances in Neural In- formation Processing Systems, vol. 28, 2015

work page 2015

[13] [13]

Regularization via mass transportation,

S. Shafieezadeh-Abadeh, D. Kuhn, and P. M. Esfahani, “Regularization via mass transportation,”Journal of Machine Learning Research, vol. 20, no. 103, pp. 1–68, 2019

work page 2019

[14] [14]

Wasserstein distributionally robust estimation in high dimensions: performance analysis and opti- mal hyperparameter tuning,

L. Aolaritei, S. Shafiee, and F. D ¨orfler, “Wasserstein distributionally robust estimation in high dimensions: performance analysis and opti- mal hyperparameter tuning,”Math. Programming, pp. 1–85, 2026

work page 2026

[15] [15]

Sequential domain adaptation by synthesizing distributionally robust experts,

B. Taskesen, M.-C. Yue, J. Blanchet, D. Kuhn, and V . A. Nguyen, “Sequential domain adaptation by synthesizing distributionally robust experts,” inInternational Conference on Machine Learning. PMLR, 2021, pp. 10 162–10 172

work page 2021

[16] [16]

Risk-aversion in multi-armed bandits,

A. Sani, A. Lazaric, and R. Munos, “Risk-aversion in multi-armed bandits,”Advances in Neural Info. Processing Systems, vol. 25, 2012

work page 2012

[17] [17]

Risk-sensitive and robust decision-making: a CVaR optimization approach,

Y . Chow, A. Tamar, S. Mannor, and M. Pavone, “Risk-sensitive and robust decision-making: a CVaR optimization approach,”Advances in Neural Information Processing Systems, vol. 28, 2015

work page 2015

[18] [18]

Risk- constrained reinforcement learning with percentile risk criteria,

Y . Chow, M. Ghavamzadeh, L. Janson, and M. Pavone, “Risk- constrained reinforcement learning with percentile risk criteria,”Jour- nal of Machine Learning Research, vol. 18, no. 167, pp. 1–51, 2018

work page 2018

[19] [19]

Fairness risk measures,

R. Williamson and A. Menon, “Fairness risk measures,” inInterna- tional Conf. on Machine Learning. PMLR, 2019, pp. 6786–6797

work page 2019

[20] [20]

Statistical learning with conditional value at risk,

T. Soma and Y . Yoshida, “Statistical learning with conditional value at risk,”arXiv preprint arXiv:2002.05826, 2020

work page arXiv 2002

[21] [21]

Superquantiles at work: Machine learning applications and efficient subgradient computation,

Y . Laguel, K. Pillutla, J. Malick, and Z. Harchaoui, “Superquantiles at work: Machine learning applications and efficient subgradient computation,”Set-Valued and Variational Analysis, vol. 29, no. 4, pp. 967–996, 2021

work page 2021

[22] [22]

Distri- butionally robust control of constrained stochastic systems,

B. P. Van Parys, D. Kuhn, P. J. Goulart, and M. Morari, “Distri- butionally robust control of constrained stochastic systems,”IEEE Transactions on Automatic Control, vol. 61, no. 2, pp. 430–442, 2015

work page 2015

[23] [23]

Toward a scalable upper bound for a CVaR-LQ problem,

M. P. Chapman and L. Lessard, “Toward a scalable upper bound for a CVaR-LQ problem,”IEEE Control Systems Letters, vol. 6, pp. 920– 925, 2021

work page 2021

[24] [24]

Risk-aware linear quadratic control using conditional value-at-risk,

M. Kishida and A. Cetinkaya, “Risk-aware linear quadratic control using conditional value-at-risk,”IEEE Transactions on Automatic Control, vol. 68, no. 1, pp. 416–423, 2022

work page 2022

[25] [25]

arXiv preprint arXiv:2112.09959 , title =

V . A. Nguyen, S. Shafiee, D. Filipovi ´c, and D. Kuhn, “Mean-covariance robust risk measurement,”arXiv preprint arXiv:2112.09959, 2021

work page arXiv 2021

[26] [26]

Distributionally robust optimization,

D. Kuhn, S. Shafiee, and W. Wiesemann, “Distributionally robust optimization,”Acta Numerica, vol. 34, p. 579–804, 2025

work page 2025

[27] [27]

Wasserstein distributionally robust optimization: Theory and appli- cations in machine learning,

D. Kuhn, P. M. Esfahani, V . A. Nguyen, and S. Shafieezadeh-Abadeh, “Wasserstein distributionally robust optimization: Theory and appli- cations in machine learning,” inOperations research & management science in the age of analytics. Informs, 2019, pp. 130–166

work page 2019

[28] [28]

A distributionally robust approach to regret optimal control using the Wasserstein distance,

F. Al Taha, S. Yan, and E. Bitar, “A distributionally robust approach to regret optimal control using the Wasserstein distance,” in62nd IEEE Conference on Decision and Control. IEEE, 2023, pp. 2768–2775

work page 2023

[29] [29]

On a formula for theL 2 Wasserstein metric be- tween measures on Euclidean and Hilbert spaces,

M. Gelbrich, “On a formula for theL 2 Wasserstein metric be- tween measures on Euclidean and Hilbert spaces,”Mathematische Nachrichten, vol. 147, no. 1, pp. 185–203, 1990

work page 1990

[30] [30]

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

P. Donti, B. Amos, and J. Z. Kolter, “Task-based end-to-end model learning in stochastic optimization,”Advances in Neural Information Processing Systems, vol. 30, 2017

work page 2017