Pith. sign in

REVIEW 2 major objections 7 minor 44 references

Decision Making under Model Misspecification: DRO with Robust Bayesian Ambiguity Sets

T0 review · 2 major / 7 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read An MMD ball around a robust nonparametric posterior predictive, instead of a KL ball around the standard posterior, keeps DRO decisions from being overly conservative under model misspecification, with a high-probability tolerance…

desk verdict Sound new construction for Bayesian DRO under misspecification; the main caveat is a tolerance guarantee that depends on the unknown misspecification gap. read the letter →

arxiv 2505.03585 v1 pith:VVHOQMSB submitted 2025-05-06 stat.ML cs.LG

Charita Dellaporta , Patrick O'Hara , Theodoros Damoulas This is my paper · ORCID
classification stat.MLcs.LG MSC 62F1590C1590C47
keywords distributionallyrobustoptimizationmodelmisspecificationmaximummeandiscrepancyBayesiannonparametriclearningNPLposteriorpredictiveambiguitysetnewsvendorproblemportfolio
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

Distributionally robust optimisation (DRO) lets a risk-averse decision maker minimize worst-case cost over an ambiguity set of distributions, and Bayesian versions center that set on the posterior. When the model is misspecified, the standard posterior drifts away from the data-generating process (DGP), so the set must be stretched to contain the truth and decisions become overly conservative. This paper claims that the problem is fixed by changing both ingredients at once: center the ambiguity set on the NPL-MMD posterior predictive, a robust posterior that puts a Dirichlet process prior directly on the DGP, and measure distance with the Maximum Mean Discrepancy (MMD), a kernel-based metric on distributions, instead of KL divergence. It proves that a radius of $C_{n,M,\alpha} + \inf_{\theta} D_k(P_\theta, P_\star)$ is enough for the set to contain the DGP with probability at least $1-\delta$, and that the true expected cost is then bounded by the worst-case cost over the set (Corollary 3.8). The resulting problem has a computable dual in a reproducing kernel Hilbert space, and Newsvendor and Portfolio experiments show better out-of-sample performance than KL-based Bayesian DRO and empirical MMD under misspecification and contamination.

What carries the argument

The load-bearing object is the pair formed by the NPL-MMD posterior predictive and the MMD ball. The predictive is $P^{\mathrm{NPL}}_{n} := E_{Q\sim \mathrm{DP}(\alpha', F')}[P_{\theta_k(Q)}]$, where $\theta_k(Q) := \arg\min_{\theta} D_k(Q, P_\theta)$ selects the model closest to a drawn DGP $Q$; the Dirichlet process prior states uncertainty directly about the DGP, and the push-forward posterior is sampled by the posterior bootstrap. The ambiguity set is $B_k^\epsilon(P) := \{P' : D_k(P', P) \le \epsilon\}$, where the MMD is a kernel-based metric on probability measures (zero only when the measures agree, for a characteristic kernel), so unlike KL balls it needs no absolute continuity. The argument runs through the mean embedding: with $C_\star := \{\mu \in \mathcal{H}_k : \|\mu - \mu_{P^{\mathrm{NPL}}_{n}}\|_k \le \epsilon\}$, the support function is $\delta^\star_{C_\star}(g) = E_{Q\sim\mathrm{DP}(\alpha',F')}\left[E_{\xi\sim P_{\theta_k(Q)}}[g(\xi)]\right] + \epsilon\|g\|_k$, and applying the kernel-DRO duality theorem turns the minimax problem into the minimization of $g_0 + \delta^\star_{C_\star}(g)$ over functions dominating $f_x$. The probabilistic tolerance guarantee is carried by a McDiarmid concentration bound on $E_{Q\sim\mathrm{DP}}[D_k(P_\star, Q)]$, which yields the radius constant $C_{n,M,\alpha}$ with rate $1/\sqrt{n}$.

What would settle it

Run repeated trials of a misspecified problem (say, a unimodal Gaussian model with a bimodal Gaussian DGP) and compare the two sides of Corollary 3.8: set $\epsilon = C_{n,M,\alpha} + \inf_\theta D_k(P_\theta, P_\star)$ and check whether the realized out-of-sample cost $E_{\xi\sim P_\star}[f_{x^\star}(\xi)]$ at the RoBAS decision exceeds $\sup_{P \in B_k^\epsilon(\hat{P}^{\mathrm{NPL}}_{n})} E_{\xi\sim P}[f_{x^\star}(\xi)]$ more often than the claimed rate $\delta$. A sharper test of Theorem 3.6 is to measure $D_k(P_\star, \hat{P}^{\mathrm{NPL}}_{n}) - \inf_\theta D_k(P_\theta, P_\star)$ as $n$ grows: if it does not decay at the claimed $1/\sqrt{n}$ rate, or if it levels off at a positive floor under bounded kernels and light-tailed DGPs, the concentration result that anchors the whole method is wrong.

Watch

Extended reading notes

Core claim

The central claim is that the NPL-MMD posterior predictive, $\hat{P}^{\mathrm{NPL}}_{n}$, inherits robustness from both its construction and its metric, so the MMD ball around it — the RoBAS ambiguity set — contains the DGP at a radius that scales with the minimal misspecification gap rather than with the posterior's sensitivity to outliers. Theorem 3.6 states that with probability at least $1-\delta$, $D_k(P_\star, \hat{P}^{\mathrm{NPL}}_{n}) \le \inf_{\theta\in\Theta} D_k(P_\theta, P_\star) + C_{n,M,\alpha}$, where $C_{n,M,\alpha}$ decays as $1/\sqrt{n}$ and is fully known given the kernel bound $M$. Corollary 3.8 then gives the tolerance guarantee: for $\epsilon \ge C_{n,M,\alpha} + \inf_{\theta} D_k(P_\theta, P_\star)$, the true expected cost is bounded above by the worst-case expectation over the set, $E_{\xi\sim P_\star}[f_x(\xi)] \le \sup_{P \in B_k^\epsilon(\hat{P}^{\mathrm{NPL}}_{n})} E_{\xi\sim P}[f_x(\xi)]$, with probability at least $1-\delta$. In the Huber contamination model the required gap is only $4\eta + 2C_{n,M,\alpha}$ (Corollary 3.9). Computationally, the set is a ball in the kernel mean embedding space, and its support function evaluates to $E_{Q\sim\mathrm{DP}(\alpha',F')}\left[E_{\xi\sim P_{\theta_k(Q)}}[g(\xi)]\right] + \epsilon\|g\|_k$, which converts the minimax problem into a finite-dimensional kernel optimization via the kernel-DRO duality theorem and the representer theorem.

Load-bearing premise

The radius guarantee only gives a high-probability bound if the decision maker knows or can closely approximate the misspecification gap $\inf_{\theta\in\Theta} D_k(P_\theta, P_\star)$ — precisely the unknown quantity the method is designed to be robust against — and the paper does not provide an estimator or data-dependent construction for it (Remark 3.7 only asks that it be 'reasonably approximated').

Editorial extensions

If this is right

  • Setting $\epsilon = C_{n,M,\alpha} + \inf_\theta D_k(P_\theta, P_\star)$ yields a decision whose true expected cost is, with probability at least $1-\delta$, no larger than the worst-case cost over the RoBAS set — so misspecification protection does not require an arbitrarily inflated ambiguity set (Corollary 3.8).
  • Because the dual problem is a kernel optimization solvable by sampling plus the representer theorem, RoBAS applies to any model family, including likelihood-free models, where DRO-BAS-PE (exponential families only) and KL-based methods needing closed-form posteriors do not apply.
  • Under Huber contamination the radius requirement grows only linearly in the contamination level ($4\eta + 2C_{n,M,\alpha}$), so mildly contaminated settings do not push decisions into extreme conservatism (Corollary 3.9).
  • In the Newsvendor and Portfolio experiments, RoBAS delivers better out-of-sample mean-variance trade-offs than DRO-BAS and BDRO under moderate-to-high misspecification, while KL-based Bayesian DRO remains competitive in the well-specified regime.
  • The duality and tolerance results are stated for a general posterior, so the RoBAS construction can be re-instantiated with other robust posteriors (generalized Bayesian inference, MMD-Bayes) without re-deriving the machinery.

Reading between the lines

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

  • The missing estimator for $\inf_\theta D_k(P_\theta, P_\star)$ could be supplied in a follow-up: for instance, a nested bootstrap on the posterior bootstrap samples (or a resampling-based upper confidence bound on the MMD gap) plugged into Remark 3.7's 'reasonable approximation' would make the radius fully data-driven, at the price of extra concentration terms; the paper stops short of that.
  • A testable corollary of Theorem 3.6 is that RoBAS's advantage over KL-based Bayesian DRO should widen as the gap between model and DGP widens: the baselines pay a radius that tracks posterior sensitivity, while RoBAS pays only the minimal MMD gap plus $C_{n,M,\alpha}$; re-running the Newsvendor experiments while sweeping a misspecification parameter would expose this scaling directly.
  • Since $C_{n,M,\alpha}$ depends on the kernel bound $M$ and the gap depends on the kernel's resolving power, kernel choice is part of the trade-off, not a free hyperparameter: a wider-bandwidth kernel shrinks the gap but inflates the constant; tuning the bandwidth against both terms is a natural extension the paper leaves open.
  • The appendix's identity $B_k^\epsilon(P^{\mathrm{NPL}}_{n}) \equiv A_k^{\epsilon^2 + v(\mathrm{DP})}$ suggests the expected-squared-MMD set carries an extra posterior-variance term $v(\mathrm{DP})$, so the RoBAS ball should be systematically less conservative than the expected-MMD set for the same nominal — an empirical comparison the paper does not report.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 7 minor

Summary. The paper proposes DRO-RoBAS, a distributionally robust optimization method whose ambiguity set is an MMD ball centered at the NPL-MMD posterior predictive, with the goal of mitigating model misspecification in Bayesian DRO. The authors derive a dual formulation in the RKHS using the kernel DRO duality theorem of Zhu et al. (2021), prove a finite-sample tolerance guarantee bounding the MMD between the data-generating process and the approximate posterior predictive by the model's irreducible MMD misspecification gap plus a concentration term, and provide an additional guarantee under Huber contamination. The method is evaluated on Newsvendor and Portfolio problems under various misspecification regimes, with comparisons against DRO-BAS, BDRO, and empirical MMD baselines.

Significance. If the claims hold, the paper makes a useful contribution by extending Bayesian DRO to settings where the parametric model is misspecified, a setting where standard Bayesian ambiguity sets can become overly conservative. The derivation is not circular: the tolerance guarantee is stated relative to the external misspecification gap inf_theta D_k(P_theta, P_star), and the dual formulation is a correct application of an existing kernel DRO duality theorem. The concentration analysis is a standard McDiarmid argument building on existing MMD posterior bootstrap results, and the experiments cover several misspecification regimes against relevant baselines. The paper also clearly identifies the computational costs and limitations of the method. The main weakness is that the central radius-selection guarantee is conditional on an unknown quantity that the paper does not operationalize, which undermines the practical actionability of the headline non-conservatism claim.

major comments (2)
  1. [Theorem 3.6, Remark 3.7, Corollary 3.8] The tolerance guarantee is not directly actionable: Corollary 3.8 requires choosing epsilon at least C_{n,M,alpha} + inf_{theta in Theta} D_k(P_theta, P_star), but inf_theta D_k(P_theta, P_star) is the unknown misspecification gap that the method is designed to be robust against. Remark 3.7 states that the gap must be 'reasonably approximated', yet no estimator, confidence bound, or data-dependent construction is provided. If the gap is underestimated, the containment guarantee and hence the upper bound on the true objective can fail; if it is overestimated, the ambiguity set is unnecessarily large and the advertised reduction in conservatism is lost. The experiments in Section 4 sweep epsilon as a free parameter, which sidesteps rather than resolves this issue. The paper should either provide an estimator for the gap with its own finite-sample guarantee, or reframe the contribution so that the radius selection problem is explicitly acknowledged as an open component.
  2. [Appendix A.2, Lemma A.2] The verification that ri(K_{C*}) is nonempty is incomplete. The proof states that 'it suffices to prove that K_{C*} is non-empty and convex', but this implication is not generally true in infinite-dimensional spaces, where a nonempty convex set can have empty relative interior. Since the dual formulation relies on Theorem 3.1 of Zhu et al. (2021), the paper should either prove directly that the MMD ball has nonempty relative interior in the relevant topology, or cite/verify a version of the duality theorem with weaker assumptions that are satisfied here.
minor comments (7)
  1. [Section 3.2] The text refers to the 'NLP-MMD posterior' in the paragraph before Assumption 3.4; this should be 'NPL-MMD posterior'.
  2. [Appendix B] The DGP specification for the contaminated Exponential case contains a duplicated and incomplete line: 'P⋆ train := (1−η)Exp(0.05) +ηN (100, 0.5) P⋆ train := (1−η)Exp(θ⋆) +ηN (µ,σ )'. This should be cleaned up.
  3. [Appendix B, Table 1 caption] The caption reads 'Average (standard derivation)solve time'; 'derivation' should be 'deviation'.
  4. [Figure 2] The legend entry 'Posterior' is ambiguous; it should be labelled 'standard posterior' to distinguish it from the NPL-MMD posterior shown in the same plot.
  5. [Appendix A.3] The concentration parameter of the DP prior is denoted alpha in the main text but c in Lemmas A.3 and A.4; the notation should be unified to avoid confusion.
  6. [Equation (17) and Theorem 3.6] The approximate DP in (17) uses a truncation level tau, but the concentration bound and the constant C_{n,M,alpha} do not explicitly depend on tau. The paper should state whether the bound is uniform in tau or clarify how tau is absorbed into the constant.
  7. [Abstract and Section 4] The abstract claims the method 'outperforms other Bayesian and empirical DRO approaches', but the experiments show that empirical MMD outperforms DRO-RoBAS in the multivariate bimodal case, and DRO-BAS formulations outperform DRO-RoBAS in the well-specified contamination case. The abstract should qualify the claim to match the paper's own conclusions.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the RoBAS derivation is self-contained, and the tolerance guarantee is conditional on an external misspecification gap rather than on the method's own fitted output.

full rationale

The paper's central derivation is not circular. The dual formulation in Corollary 3.3 follows from the external dual theorem of Zhu et al. (2021) applied to the MMD ball C*, with the support function computed directly from the definition of C* in Proposition 3.2 and Lemma A.1; no target result is assumed in the construction. The tolerance guarantee in Theorem 3.6 bounds D_k(P*, hat P^NPL_pred_n) by inf_theta D_k(P_theta,P*) plus a finite-sample constant, using a concentration lemma adapted from Dellaporta et al. (2022). That cited result is an independent published technical bound with stated assumptions (bounded kernel, DP approximation scheme) and does not assume the paper's intended conclusion, so the self-citation is legitimate support rather than circularity. The theorem is admittedly conditional on the unknown misspecification gap: Remark 3.7 states that 'if inf_{theta in Theta} D_k(P_theta, P_star) can be reasonably approximated, this result can be used to select the radius ensuring RoBAS includes P_star with high probability.' This is an actionability limitation, not a circular reduction, because the gap is an external property of the model family and DGP, not the output of the RoBAS optimization or a parameter fitted to the same prediction the paper claims. No equation in the paper is shown to be definitionally equivalent to its own conclusion, and no fitted input is renamed as a prediction. Accordingly, no circular step is identified; the paper's contribution remains substantive even though the guaranteed radius is not directly operationalizable without additional estimation of the misspecification gap.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The central claim rests on a handful of standard results (kernel DRO duality, McDiarmid's inequality, DP posterior) and two domain assumptions (bounded kernel, bounded sublevel sets). The most notable dependence is on Lemmas 6 and 11 of Dellaporta et al. (2022), a prior paper with overlapping authors, which are used as black boxes in the proof of Theorem 3.6; these lemmas are peer-reviewed but not re-derived here. The method itself uses a concentration parameter alpha, a kernel bandwidth, and the DRO radius epsilon as user-chosen inputs rather than fitted parameters.

free parameters (3)
  • epsilon (DRO radius) = not fitted; swept over a grid in experiments
    Radius of the MMD ambiguity set. The tolerance guarantee relates it to the unknown misspecification gap inf_theta D_k(P_theta, P_star), so it is not prescribed without additional estimation.
  • kernel bandwidth h = median heuristic on data
    Bandwidth of the Gaussian kernel used for MMD. Chosen by the median heuristic (Gretton et al. 2012), not optimized for the DRO objective, but it changes the geometry of the ambiguity set.
  • DP concentration alpha = alpha = 0 in experiments
    Concentration parameter of the Dirichlet process prior on the DGP. Set to 0, a non-informative prior, in all experiments, which makes the posterior centered on the empirical measure; the theory allows alpha >= 0.
assumptions (6)
  • standard math Theorem 3.1 of Zhu et al. (2021), duality for kernel DRO
    The dual formulation of DRO-RoBAS (Corollary 3.3) is a direct application of this theorem to the MMD ball C*. The theorem requires f proper, upper semicontinuous, and non-empty relative interior of the ambiguity set.
  • domain assumption Bounded kernel (Assumption 3.5): |k(xi,xi')| <= M
    Needed for the McDiarmid concentration inequality and the bound on expected MMD; satisfied by the Gaussian kernel used in experiments.
  • domain assumption Bounded sublevel sets for theta_k(Q) (Assumption 3.4)
    Ensures existence of a minimizer in the NPL fitting step (8); standard in minimum MMD estimation.
  • standard math Lemmas 6 and 11 of Dellaporta et al. (2022) bounding E[D(P_star, Pn)] and E[D(Pn, Q)]
    Used directly in Lemma A.3 and hence in Theorem 3.6. These are results from a prior paper with overlapping authors; they are peer-reviewed but not re-derived here.
  • domain assumption DP posterior concentration (5) and approximate DP sampling (17)
    Define the posterior over the DGP; the approximation (17) is the practical object for which Theorem 3.6 is proved.
  • standard math McDiarmid's inequality
    Used in Lemma A.4 to convert the expected MMD bound into a high-probability bound.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Decision Making under Model Misspecification: DRO with Robust Bayesian Ambiguity Sets." pith.science (2026). https://pith.science/paper/VVHOQMSB

@misc{pith2026250503585,
  author       = {Pith},
  title        = {Pith review of: Decision Making under Model Misspecification: DRO with Robust Bayesian Ambiguity Sets},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VVHOQMSB}},
  note         = {Machine review of arXiv:2505.03585}
}
read the original abstract

Distributionally Robust Optimisation (DRO) protects risk-averse decision-makers by considering the worst-case risk within an ambiguity set of distributions based on the empirical distribution or a model. To further guard against finite, noisy data, model-based approaches admit Bayesian formulations that propagate uncertainty from the posterior to the decision-making problem. However, when the model is misspecified, the decision maker must stretch the ambiguity set to contain the data-generating process (DGP), leading to overly conservative decisions. We address this challenge by introducing DRO with Robust, to model misspecification, Bayesian Ambiguity Sets (DRO-RoBAS). These are Maximum Mean Discrepancy ambiguity sets centred at a robust posterior predictive distribution that incorporates beliefs about the DGP. We show that the resulting optimisation problem obtains a dual formulation in the Reproducing Kernel Hilbert Space and we give probabilistic guarantees on the tolerance level of the ambiguity set. Our method outperforms other Bayesian and empirical DRO approaches in out-of-sample performance on the Newsvendor and Portfolio problems with various cases of model misspecification.

Figures

Figures reproduced from arXiv: 2505.03585 by the authors.

Figure 1
Figure 1. Illustration of (approximated) BASPE, BASPP and RoBAS (ours) with Pθ = N (µ, σ2 ) over a grid of (µ, σ) pairs for a fixed ϵ. In the well-specified case (left), all ambiguity sets include the DGP while RoBAS covers a slightly bigger area than BASPE and BASPP. For a contaminated dataset (right) RoBAS con￾tinues to contain the DGP and maintains a similar area, whereas the BAS formulations exclude it and cover a much la… view at source ↗
Figure 2
Figure 2. Contaminated Gaussian location example. Top: His￾togram of observed data along with the true (DGP), outlier and pathological densities. Bottom: Posterior marginal distributions for NPL-MMD and standard Bayes. The true mean is indicated with a dotted line. For BASPE it holds that EΠBayes [dKL(Ppathological, Pθ)] ≈ 0.17 < 0.42 ≈ EΠBayes [dKL(P ⋆ , Pθ)] and similarly for BASPP it holds that dKL(Ppathological, P pred n … view at source ↗
Figure 3
Figure 3. The out-of-sample mean and variance for the Newsven￾dor problem with a misspecified Gaussian location model and a bi￾modal Gaussian DGP. Results are shown for the univariate (D = 1, top) and the multivariate (D = 5, bottom) cases, with markers representing ϵ values. For illustration purposes, the bottom-left area of the multivariate case is shown in a zoomed-in view. in Staib & Jegelka (2019) and also forms a specia… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Out-of-sample mean-variance trade-off in the Newsvendor problem for a Gaussian location model (top) and an Exponential model (bottom) with a contaminated training dataset. Results are shown for contamination levels η = 0.0 (left), η = 0.1 (middle), and η = 0.2 (right).…
Figure 5
Figure 5. Figure 5: Out-of-sample mean-variance trade-off in the Portfolio problem for a 5D contaminated Gaussian DGP with η = 0.0, 0.1, 0.2. Note that the goal is to maximise returns, so larger out-of-sample mean is better. 4.2. The Portfolio Optimisation Problem We continue with the mul…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 18 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    and Gerber, M

    Alquier, P. and Gerber, M. Universal robust regression via maximum mean discrepancy. Biometrika, 111 0 (1): 0 71--92, 2024

    work page 2024

  3. [3]

    Robust and conjugate G aussian process regression

    Altamirano, M., Briol, F.-X., and Knoblauch, J. Robust and conjugate G aussian process regression. arXiv preprint arXiv:2311.00463, 2023

    arXiv 2023

  4. [4]

    Sharp analysis of low-rank kernel matrix approximations

    Bach, F. Sharp analysis of low-rank kernel matrix approximations. In Conference on learning theory, pp.\ 185--209. PMLR, 2013

    2013

  5. [5]

    G., Holmes, C

    Bissiri, P. G., Holmes, C. C., and Walker, S. G. A general framework for updating belief distributions. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 78 0 (5): 0 1103--1130, 2016

    2016

  6. [6]

    Parametric distributionally robust optimisation models for resource and inventory planning problems

    Black, B. Parametric distributionally robust optimisation models for resource and inventory planning problems. Lancaster University (United Kingdom), 2022

    work page 2022

  7. [7]

    B., and Girolami, M

    Briol, F.-X., Barp, A., Duncan, A. B., and Girolami, M. Statistical inference for generative models with maximum mean discrepancy. arXiv preprint arXiv:1906.05944, 2019

    arXiv 1906

  8. [8]

    Distributionally robust decision making leveraging conditional distributions

    Chen, Y., Kim, J., and Anderson, J. Distributionally robust decision making leveraging conditional distributions. In 2022 IEEE 61st Conference on Decision and Control (CDC), pp.\ 5652--5659. IEEE, 2022

    work page 2022

Show all 44 references
  1. [9]

    and Alquier, P

    Ch \'e rief-Abdellatif, B.-E. and Alquier, P. MMD - B ayes: Robust B ayesian estimation via maximum mean discrepancy. In Symposium on Advances in Approximate Bayesian Inference, pp.\ 1--21. PMLR, 2020

    2020

  2. [10]

    and Alquier, P

    Ch \'e rief-Abdellatif, B.-E. and Alquier, P. Finite sample properties of parametric MMD estimation: robustness to misspecification and dependence. Bernoulli, 28 0 (1): 0 181--213, 2022

    2022

  3. [11]

    Robust B ayesian inference for simulator-based models via the MMD posterior bootstrap

    Dellaporta, C., Knoblauch, J., Damoulas, T., and Briol, F.-X. Robust B ayesian inference for simulator-based models via the MMD posterior bootstrap. In International Conference on Artificial Intelligence and Statistics, pp.\ 943--970. PMLR, 2022

    2022

  4. [12]

    Decision making under the exponential family: Distributionally robust optimisation with bayesian ambiguity sets

    Dellaporta, C., O'Hara, P., and Damoulas, T. Decision making under the exponential family: Distributionally robust optimisation with bayesian ambiguity sets. arXiv preprint arXiv:2411.16829, 2024

    2024 arXiv

  5. [13]

    Scalable nonparametric sampling from multimodal posteriors with the posterior bootstrap

    Fong, E., Lyddon, S., and Holmes, C. Scalable nonparametric sampling from multimodal posteriors with the posterior bootstrap. In International Conference on Machine Learning, pp.\ 1952--1962. PMLR, 2019

    1952

  6. [14]

    and Kleywegt, A

    Gao, R. and Kleywegt, A. Distributionally robust stochastic optimization with W asserstein distance. Mathematics of Operations Research, 48 0 (2): 0 603--655, 2023

    2023

  7. [15]

    and Basu, A

    Ghosh, A. and Basu, A. Robust B ayes estimation using the density power divergence. Annals of the Institute of Statistical Mathematics, 68: 0 413--437, 2016

    2016

  8. [16]

    M., Rasch, M

    Gretton, A., Borgwardt, K. M., Rasch, M. J., Sch \"o lkopf, B., and Smola, A. A kernel two-sample test. The Journal of Machine Learning Research, 13 0 (1): 0 723--773, 2012

    2012

  9. [17]

    The safe B ayesian: learning the learning rate via the mixability gap

    Gr \"u nwald, P. The safe B ayesian: learning the learning rate via the mixability gap. In International Conference on Algorithmic Learning Theory, pp.\ 169--183. Springer, 2012

    2012

  10. [18]

    and Hong, L

    Hu, Z. and Hong, L. J. Kullback- L eibler divergence constrained distributionally robust optimization. Available at Optimization Online, 1 0 (2): 0 9, 2013

    2013

  11. [19]

    Huber, P. J. Robust estimation of a location parameter. In Breakthroughs in statistics: Methodology and distribution, pp.\ 492--518. Springer, 1992

    1992

  12. [20]

    Husain, H., Nguyen, V., and Hengel, A. v. d. Distributionally robust B ayesian optimization with -divergences. arXiv preprint arXiv:2203.02128, 2022

    2022 arXiv

  13. [21]

    Hedging against complexity: D istributionally robust optimization with parametric approximation

    Iyengar, G., Lam, H., and Wang, T. Hedging against complexity: D istributionally robust optimization with parametric approximation. In International Conference on Artificial Intelligence and Statistics, pp.\ 9976--10011. PMLR, 2023

    2023

  14. [22]

    Q., and Holmes, C

    Jewson, J., Smith, J. Q., and Holmes, C. Principles of B ayesian inference using general divergence criteria. Entropy, 20 0 (6): 0 442, 2018

    2018

  15. [23]

    Kingma, D. P. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  16. [24]

    An optimization-centric view on B ayes' rule: Reviewing and generalizing variational inference

    Knoblauch, J., Jewson, J., and Damoulas, T. An optimization-centric view on B ayes' rule: Reviewing and generalizing variational inference. Journal of Machine Learning Research, 23 0 (132): 0 1--109, 2022

    2022

  17. [25]

    M., Nguyen, V

    Kuhn, D., Esfahani, P. M., Nguyen, V. A., and Shafieezadeh-Abadeh, S. Wasserstein distributionally robust optimization: Theory and applications in machine learning. In Operations Research & Management Science in the Age of Analytics, chapter 6, pp.\ 130--166. Informs, 2019. do...

    2019

  18. [26]

    Li, J. Y. and Kwon, R. H. Portfolio selection under model uncertainty: a penalized moment-based optimization approach. Journal of Global Optimization, 56: 0 131--164, 2013

    2013

  19. [27]

    Towards out-of-distribution generalization: A survey

    Liu, J., Shen, Z., He, Y., Zhang, X., Xu, R., Yu, H., and Cui, P. Towards out-of-distribution generalization: A survey. arXiv preprint arXiv:2108.13624, 2021

    2021 arXiv

  20. [28]

    Lyddon, S., Walker, S., and Holmes, C. C. Nonparametric learning from B ayesian models with randomized objective functions. Advances in neural information processing systems, 31, 2018

    2018

  21. [29]

    Matsubara, T., Knoblauch, J., Briol, F.-X., and Oates, C. J. Robust generalised B ayesian inference for intractable likelihoods. Journal of the Royal Statistical Society Series B: Statistical Methodology, 84 0 (3): 0 997--1022, 2022

    2022

  22. [30]

    McDiarmid, C. et al. On the method of bounded differences. Surveys in combinatorics, 141 0 (1): 0 148--188, 1989

    1989

  23. [31]

    Modeling the second player in distributionally robust optimization

    Michel, P., Hashimoto, T., and Neubig, G. Modeling the second player in distributionally robust optimization. arXiv preprint arXiv:2103.10282, 2021

    2021 arXiv

  24. [32]

    Distributionally robust models with parametric likelihood ratios

    Michel, P., Hashimoto, T., and Neubig, G. Distributionally robust models with parametric likelihood ratios. arXiv preprint arXiv:2204.06340, 2022

    2022 arXiv

  25. [33]

    Integral probability metrics and their generating classes of functions

    M \"u ller, A. Integral probability metrics and their generating classes of functions. Advances in applied probability, 29 0 (2): 0 429--443, 1997

    1997

  26. [34]

    Porteus, E. L. Stochastic inventory theory. Handbooks in operations research and management science, 2: 0 605--652, 1990

    1990

  27. [35]

    and Recht, B

    Rahimi, A. and Recht, B. Random features for large-scale kernel machines. Advances in neural information processing systems, 20, 2007

    2007

  28. [36]

    and Mehrotra, S

    Rahimian, H. and Mehrotra, S. Distributionally robust optimization: A review. arXiv preprint arXiv:1908.05659, 2019

    1908 arXiv

  29. [37]

    R., and Abate, A

    Romao, L., Hota, A. R., and Abate, A. Distributionally robust optimal and safe control of stochastic systems via kernel conditional mean embedding. In 2023 62nd IEEE Conference on Decision and Control (CDC), pp.\ 2016--2021. IEEE, 2023

    2023

  30. [38]

    Bayesian distributionally robust optimization

    Shapiro, A., Zhou, E., and Lin, Y. Bayesian distributionally robust optimization. SIAM Journal on Optimization, 33 0 (2): 0 1279--1304, 2023

    2023

  31. [39]

    and Jegelka, S

    Staib, M. and Jegelka, S. Distributionally robust optimization and generalization in kernel methods. Advances in Neural Information Processing Systems, 32, 2019

    2019

  32. [40]

    G., R \"u ckmann, J.-J., Stein, O., and Still, G

    V \'a zquez, F. G., R \"u ckmann, J.-J., Stein, O., and Still, G. Generalized semi-infinite programming: a tutorial. Journal of computational and applied mathematics, 217 0 (2): 0 394--419, 2008

    2008

  33. [41]

    Walker, S. G. Bayesian inference with misspecified models. Journal of statistical planning and inference, 143 0 (10): 0 1621--1633, 2013

    2013

  34. [42]

    Learning against distributional uncertainty: On the trade-off between robustness and specificity

    Wang, S., Wang, H., and Honorio, J. Learning against distributional uncertainty: On the trade-off between robustness and specificity. arXiv preprint arXiv:2301.13565, 2023

    2023 arXiv

  35. [43]

    Coping with label shift via distributionally robust optimisation

    Zhang, J., Menon, A., Veit, A., Bhojanapalli, S., Kumar, S., and Sra, S. Coping with label shift via distributionally robust optimisation. arXiv preprint arXiv:2010.12230, 2020

    2010 arXiv

  36. [44]

    Kernel distributionally robust optimization: G eneralized duality theorem and stochastic approximation

    Zhu, J.-J., Jitkrittum, W., Diehl, M., and Sch \"o lkopf, B. Kernel distributionally robust optimization: G eneralized duality theorem and stochastic approximation. In International Conference on Artificial Intelligence and Statistics, pp.\ 280--288. PMLR, 2021

    2021

Pith tools

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