Pith. sign in

REVIEW 4 major objections 5 minor 40 references

Distributionally Faithful Imputation via Positive Semi-Definite Kernel Density Estimation

T0 review · 4 major / 5 minor · reviewed 2026-07-10 · grok-4.5

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

Pith's one-line read Imputation under MCAR is convex density estimation that matches observed marginals, solved with PSD kernel models that give both single and multiple fills from one fit.

desk verdict Solid convex PSD-imputation method with real theory and competitive early experiments; the 'minimax' rate claim is overstated by a reverse-Pinsker balancing factor. read the letter →

arxiv 2607.07767 v1 pith:FKWRTKYP submitted 2026-07-08 stat.ML cs.LG

Andrea Basteri , Carlo Ciliberto , Alessandro Rudi This is my paper · ORCID
classification stat.MLcs.LG MSC 62G0762D1068T05
keywords missingdataMCARimputationPSDkerneldensitiesmarginalmatchingconvexoptimisationdistributionalfidelitymultiple
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

Missing values break statistical pipelines, and most imputers either chase pointwise errors or lean on restrictive models that never recover a coherent joint. This paper reframes MCAR imputation as estimating a density whose observed marginals match those of the data. Positive semi-definite kernel densities make the resulting empirical risk convex, supply closed-form marginals, and admit a Newton interior-point solver. One fitted density then produces both single imputations (conditional means) and multiple imputations (samples). Theory shows the estimator is consistent for the observable marginals at an adaptive rate that can beat the usual curse of dimensionality when the true density is sufficiently smooth. Early experiments on synthetic manifolds and eleven real tables already match or beat strong baselines on distributional metrics while remaining competitive on RMSE.

What carries the argument

PSD kernel densities: non-negative functions p_Q(x) = φ(x)^T Q φ(x) with Q ≽ 0 and a closed-form normalising constraint, whose every marginal is again an explicit trace formula. This structure turns the masked negative log-likelihood into a convex optimisation problem over the PSD cone.

What would settle it

On a compact domain where the true density is only Lipschitz or has a non-manifold zero set, measure whether the observed excess masked KL still decays at the claimed rate as sample size grows; a clear plateau or slower rate would refute the consistency theorem.

Watch

Extended reading notes

Core claim

Under MCAR, the only recoverable object is the family of observed marginals. Minimising the masked Kullback–Leibler risk over the class of PSD kernel densities yields a convex program whose solution is statistically consistent for those marginals and, from the same density, supplies both single and multiple imputations.

Load-bearing premise

The true density must be Hölder-smooth of order higher than half the dimension and either stay bounded away from zero or have zeros that form a smooth manifold; without that regularity the approximation rates and the conversion from L2 error into KL control fail.

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

4 major / 5 minor

Summary. The paper reframes MCAR imputation as estimating a density whose observed marginals match those of the data, using positive semi-definite (PSD) kernel densities. This yields a convex regularized empirical-risk problem (Eq. 4 / Problem P) with closed-form marginals, solved by a damped Newton interior-point method. The same fitted density supports both single (conditional mean) and multiple imputation. Under Hölder smoothness and mesh assumptions, the authors prove consistency of the masked KL risk, culminating in Theorem 5.4’s rate O(N^{-s/(6s+2d)}). Preliminary experiments on synthetic manifolds and eleven real tabular datasets report competitive energy distance and OT scores against mean, IterativeImputer, SoftImpute, and OT-Impute baselines.

Significance. If the technical claims hold, the work offers a rare combination of (i) a convex, distributionally motivated objective for imputation, (ii) closed-form marginals that make the ERM tractable, and (iii) a single coherent density for both single and multiple imputation—properties that joint-model, FCS, low-rank, and deep generative methods rarely share. Full proofs of convexity (Thm 4.1), approximation (Thm 5.1–5.2), generalization (Thm 5.3), and the combined rate (Thm 5.4) appear in Appendices B–C and rest on standard tools (Rademacher complexity, reverse Pinsker, PSD approximation theory). That package is a genuine contribution to principled missing-data methodology. The practical significance is currently limited by the preliminary experimental scale and by overstated rate claims that need correction before the theoretical contribution can be fairly assessed.

major comments (4)
  1. [Theorem 5.4 / Abstract / §1] Abstract, §1 Contributions (3), and the discussion around Theorem 5.4 repeatedly describe the excess-risk rate as “classical minimax,” “O(1/√N),” or “beating the curse of dimensionality for very regular probabilities.” Theorem 5.4 actually establishes ∫ KL(p⋆_S ∥ p̂_S) dμ(S) = O(N^{-s/(6s+2d)}). The classical nonparametric KL/minimax rate for Hölder-s densities is O(N^{-s/(2s+d)}); the paper’s own L² approximation theory (Thm 5.1 / B.1) would support that order if the excess-risk analysis closed at the same scale. The final exponent is degraded by roughly a factor of three. For large s the stated rate tends to N^{-1/6}, not N^{-1/2}. These statements must be corrected or the stronger rate proved; as written they overstate the result.
  2. [§5.1 Assumption (i), Theorem 5.2, Appendix B.3] The degradation is forced by the reverse-Pinsker step in the proof of Theorem 5.2: L((1−ν)p_Q̄ + νu) ≤ log(C₂/ν)·∥(1−ν)p_Q̄ + νu − p⋆∥₁. That step requires Assumption (i) (p⋆ bounded away from zero, or zeros forming a C¹ manifold) and introduces the log(1/ν) factor that forces the suboptimal balancing ν ∼ N^{-s/(6s+2d)}. If the lower-bound assumption fails even mildly, both the rate and the claim of minimax optimality for the observable marginals become unsupported. The manuscript should either strengthen the KL control (e.g., via a different inequality or a truncated risk) or clearly label the rate as suboptimal and state the assumption’s necessity more prominently in the main text.
  3. [Appendix A / Abstract / §8] Section A states that the implementation uses α = 0 in Problem P, that anchor budgets are capped at 65–85, that hyper-parameters are tuned by a two-stage CV + alternating minimization, and that results are “preliminary.” The abstract and conclusion nevertheless claim “competitive distributional accuracy” and “strong practical promise.” With α = 0 the log term is unbounded below when Tr(QA_i) = 0, the reported variance across seeds is non-negligible, and the method is restricted to d ≲ 60, n ≲ 1.6k. Either the experimental claims should be tempered to match the exploratory status, or a more complete benchmark (including α > 0, larger d, and runtime/memory profiles consistent with §6.1) should be supplied.
  4. [§1 Contributions / Theorem 5.3] Contribution (3) in §1 lists “an O(1/√N) excess-risk bound” separately from the consistency statement. Theorem 5.3 does give a generalization term of order (ℓ + λ + 1/λ) log^{3/2}(·)/√N, but after balancing with approximation error the final masked-KL rate is the slower quantity of Theorem 5.4. Presenting O(1/√N) as a headline guarantee without the balancing caveat is misleading; the contribution list should be aligned with the theorems that are actually proved.
minor comments (5)
  1. [§3–§4] Notation for the mask distribution switches between μ(S) and m(S); the feature map ϕ and the moment matrix H are redefined with slightly different domains in §4 and §6. A single consistent notation block would help.
  2. [Figure 2] Figure 2 caption claims PSD-Impute “outperforms baselines by ≥15% on distributional metrics”; the plotted bars do not uniformly support a 15% relative improvement across all datasets and missing rates. Soften or quantify per-panel.
  3. [Abstract / §1] Several references to “classical minimax” and “beating the curse” appear before any rate is stated; a forward pointer to Theorem 5.4 (with the corrected wording) would avoid early over-claim.
  4. [Throughout / Appendix D] Typos: “regulirized” (§4), “dimen-sions” (§5.5), “T r” vs “Tr” inconsistency in Appendix D, and “ice_mi” in Figure 5 caption (presumably IterativeImputer MI).
  5. [§6.1] The complexity derivation in §6.1 quotes memory O(ε^{-24−10d/s}) for the SWM method; a short sanity check against the moderate-d regime used in experiments would make the asymptotic claim more credible.

Circularity Check

1 steps flagged · score 2.0 of 10

Minor non-load-bearing self-citation of PSD approximation theory from overlapping authors; the masked-risk ERM, learning-error bounds, and consistency derivation are independently developed and do not reduce to inputs by construction.

  1. self citation load bearing [§5.3 Theorem 5.1 and its proof sketch; also §4 (i)–(iv) and Appendix B.1–B.2]
    "The theory of PSD models [26] tells us under assumption (i) it is possible to efficiently approximate a probability with a small mesh W, i.e. that there exists a good candidate Q̄ i R^{ℓ imesℓ}, Q̄≽0, such that p_Q̄ approximates very effectively p∗ and with a small Frobenius norm for Q̄. … When ℓ≥C1 ϵ^{-d/s}(log(C1/ϵδ))^{d/2}, η=C2 ϵ^{-2/s}/log(2/ϵ), then there exists Q̄ … ∥p_Q̄-p∗∥_{L2(Ω)}≤ϵ, tr(Q̄K)≤C3ℓ."

    The existence of a low-complexity PSD approximant that underpins both the approximation-error bound (Thm 5.2) and the final rate (Thm 5.4) is taken from the authors' own prior paper [26] rather than re-derived from first principles. The citation is not circular in the strong sense (the prior theorem is independent of the imputation setting and supplies a genuine mathematical fact under the stated Hölder assumptions), yet it is the sole source of the approximation half of the excess-risk decomposition; without it the claimed rate would not follow. This is therefore a mild, non-load-bearing self-citation of the kind that scores 2.

full rationale

The paper's derivation chain is standard statistical learning: define masked KL risk L(p) whose zeros are exactly the distributions matching the observable marginals (Lemma 3.1), replace by the empirical negative log-likelihood of the masked observations (eq. 3), restrict to the PSD model class whose closed-form marginals make the problem convex (Thm 4.1), then bound excess risk by the usual approximation + generalization decomposition (eq. 6). The approximation half (Thm 5.1/B.1) imports the existence of a good PSD approximant from the authors' prior work [26], but that prior result is a general, parameter-free approximation theorem for Hölder densities under stated assumptions that do not mention imputation or the masked risk; the learning-error half (Thm 5.3), the balancing that produces the final rate (Thm 5.4), the interior-point solver, and the experimental protocol are all new and self-contained. No quantity is fitted to data and then re-presented as a prediction, no uniqueness theorem is imported to forbid alternatives, and the risk itself is not defined in terms of the estimator. The self-citation therefore raises the score only to the mild level of 2; the central consistency claim does not collapse by construction.

Assumptions & free parameters 5 free parameters · 5 assumptions · 1 invented entities

The central claim rests on the classical MCAR factorization, on Hölder regularity of p⋆, on the approximation power of PSD kernel densities from prior work, and on a collection of free regularization and mesh parameters that are chosen by CV/alternating minimization rather than fixed by theory. No new physical entities are postulated; the PSD-Impute model is a constructed estimator, not an invented ontological object.

free parameters (5)
  • regularization λ (RKHS / trace term)
    Controls complexity of Q; theory gives a rate-dependent schedule, practice uses CV and alternating minimization.
  • mixture weight ν (uniform floor)
    Appears in the smoothed density (1−ν)p_Q + νu used for reverse Pinsker and generalization; chosen as Õ(N^{-s/(6s+2d)}) in theory, free in practice.
  • barrier weight μ and schedule
    Self-concordant barrier for the interior-point Newton method; decreased geometrically; selected from a discrete grid by CV.
  • kernel precision η and anchor set W (size ℓ)
    Representation hyperparameters; ℓ capped at 65–85 in experiments; W initialized from an imputed copy and moved by alternating minimization / Adam.
  • smoothing α in log(Tr(QA_i)+α)
    Makes the loss upper-bounded; experiments set α=0, which is a special case of the analyzed problem.
assumptions (5)
  • domain assumption Missingness is MCAR: mask S is independent of X, so the joint factors as p⋆(x)⊗μ(S) (Section 3).
    Load-bearing for the claim that matching observed marginals is the right population target; MAR is only mentioned as known to work for MLE, not analyzed.
  • domain assumption p⋆ is Hölder-s on a compact domain with s>d/2, bounded above, and either bounded away from zero or with zeros a C¹ manifold (Assumption 5.1(i)).
    Required for PSD approximation rates and for reverse Pinsker conversion of L² error into KL.
  • standard math PSD kernel densities with Gaussian kernels have closed-form marginals and are dense in L¹ as anchors densify (properties from [17,26]).
    Imported as black-box representation theory; used throughout Sections 4–5.
  • domain assumption Anchor mesh W is drawn uniformly i.i.d. on the hypercube (Assumption 5.1(ii)) so fill distance concentrates.
    Used to invoke the existence of a good PSD approximant (Thm 5.1); practice uses data-dependent heuristics instead.
  • ad hoc to paper KL is the right Bregman divergence for the empirical proxy (log-likelihood) of the masked risk (Section 3).
    Motivated by information theory and convexity, but other divergences (e.g., OT) are also used as evaluation metrics; the training objective choice is a modeling decision.
invented entities (1)
  • PSD-Impute estimator (p_Q with Q from regularized masked ERM)
    purpose: Concrete convex model that produces single/multiple imputations from one density.
    Constructed estimator, not a new physical object; independent evidence is the theory + preliminary experiments, not an external measurement.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Distributionally Faithful Imputation via Positive Semi-Definite Kernel Density Estimation." pith.science (2026). https://pith.science/paper/FKWRTKYP

@misc{pith2026260707767,
  author       = {Pith},
  title        = {Pith review of: Distributionally Faithful Imputation via Positive Semi-Definite Kernel Density Estimation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FKWRTKYP}},
  note         = {Machine review of arXiv:2607.07767}
}
read the original abstract

Missing values undermine statistical inference and machine learning pipelines, yet most imputation methods rely on heuristics or restrictive parametric assumptions that ignore the joint data distribution. We recast imputation under missing completely at random (MCAR) as density estimation from masked observations: estimate a distribution whose observed marginals exactly match those in the data. Leveraging positive semi definite (PSD) kernel densities we obtain a convex empirical risk problem with closed form marginals, solvable by a Newton interior point method. The resulting PSD Impute model yields both single and multiple imputations from the same fitted density, enjoys statistical consistency with fast adaptive excess risk beating the curse of dimensionality for very regular probabilities. Preliminary experiments on one synthetic and eleven real world datasets already indicate competitive distributional accuracy compared with popular imputation baselines, suggesting strong practical promise.

Figures

Figures reproduced from arXiv: 2607.07767 by the authors.

Figure 1
Figure 1. Qualitative results on toy datasets. Green: input points (missing coordinates projected), [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Distributional accuracy on real datasets. Top: Optimal Transport (OT); bottom: Energy Distance (ED). Columns correspond to missing rates p = 0.2 (left) and 0.4 (right). PSD-Impute outperforms baselines by ≥15% on distributional metrics while matching RMSE. Metrics. Energy Distance (ED) [31] is our primary criterion because it compares full joint distribu￾tions; 2–Wasserstein OT (appendix) and RMSE [20] are reported … view at source ↗
Figure 3
Figure 3. RMSE counterparts to Figure 2. Lower is better. As expected, methods tuned for RMSE [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative results on 2–D (top) and 3–D (bottom) toy datasets. Top: input points (missing [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: Energy Distance (multiple imputation). PSD multiple imputation matches or surpasses [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 40 canonical work pages

  1. [1]

    Springer Science & Business Media, 2013

    Rajendra Bhatia.Matrix analysis. Springer Science & Business Media, 2013

    work page 2013

  2. [2]

    Jeff A Bilmes et al. A gentle tutorial of the em algorithm and its application to parameter estimation for gaussian mixture and hidden markov models.International computer science institute, 4(510):126, 1998

    work page 1998

  3. [3]

    A note on reverse pinsker inequalities.IEEE transactions on information theory, 65(7):4094–4096, 2019

    Olivier Binette. A note on reverse pinsker inequalities.IEEE transactions on information theory, 65(7):4094–4096, 2019

    work page 2019

  4. [4]

    Cambridge University Press, 2004

    Stephen Boyd and Lieven Vandenberghe.Convex Optimization. Cambridge University Press, 2004

    work page 2004

  5. [5]

    Alternating minimization as sequential unconstrained minimization: a survey

    Charles L Byrne. Alternating minimization as sequential unconstrained minimization: a survey. Journal of Optimization Theory and Applications, 156:554–566, 2013

    work page 2013

  6. [6]

    John Wiley & Sons, 1999

    Thomas M Cover.Elements of information theory. John Wiley & Sons, 1999. 9

    work page 1999

  7. [7]

    I-divergence geometry of probability distributions and minimization problems

    Imre Csiszár. I-divergence geometry of probability distributions and minimization problems. The Annals of Probability, 3(1):146–158, 1975

    work page 1975

  8. [8]

    Maximum likelihood from incomplete data via the em algorithm.Journal of the Royal Statistical Society: Series B (Methodological), 39(1):1–38, 1977

    Arthur P Dempster, Nan M Laird, and Donald B Rubin. Maximum likelihood from incomplete data via the em algorithm.Journal of the Royal Statistical Society: Series B (Methodological), 39(1):1–38, 1977

    work page 1977

Show all 40 references
  1. [9]

    Matrix completion and low-rank svd via fast alternating least squares.Journal of Machine Learning Research, 16:3367–3402, 2015

    Trevor Hastie, Rahul Mazumder, Jason Lee, and Reza Zadeh. Matrix completion and low-rank svd via fast alternating least squares.Journal of Machine Learning Research, 16:3367–3402, 2015

    2015

  2. [10]

    On the consistency of supervised learning with missing values.Statistical Papers, 65(9):5447–5479, 2024

    Julie Josse, Jacob M Chen, Nicolas Prost, Gaël Varoquaux, and Erwan Scornet. On the consistency of supervised learning with missing values.Statistical Papers, 65(9):5447–5479, 2024

    2024

  3. [11]

    Regularization techniques for learning with matrices.The Journal of Machine Learning Research, 13(1):1865–1890, 2012

    Sham M Kakade, Shai Shalev-Shwartz, and Ambuj Tewari. Regularization techniques for learning with matrices.The Journal of Machine Learning Research, 13(1):1865–1890, 2012

    2012

  4. [12]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. InInterna- tional Conference on Learning Representations, 2015

    2015

  5. [13]

    Imputation for prediction: beware of diminishing returns.arXiv preprint arXiv:2407.19804, 2024

    Marine Le Morvan and Gaël Varoquaux. Imputation for prediction: beware of diminishing returns.arXiv preprint arXiv:2407.19804, 2024

    2024 arXiv

  6. [14]

    John Wiley and Sons, 3 edition, 2019

    Roderick J A Little and Donald B Rubin.Statistical Analysis with Missing Data. John Wiley and Sons, 3 edition, 2019

    2019

  7. [15]

    Explicit and recursive estimates of the lambert w function.arXiv preprint arXiv:2008.06122, 2020

    Lajos Lóczi. Explicit and recursive estimates of the lambert w function.arXiv preprint arXiv:2008.06122, 2020

    2008 arXiv

  8. [16]

    Vaem: a deep generative model for heterogeneous mixed type data.Advances in Neural Information Processing Systems, 33:11237–11247, 2020

    Chao Ma, Sebastian Tschiatschek, Richard Turner, José Miguel Hernández-Lobato, and Cheng Zhang. Vaem: a deep generative model for heterogeneous mixed type data.Advances in Neural Information Processing Systems, 33:11237–11247, 2020

    2020

  9. [17]

    Non-parametric models for non- negative functions.Advances in neural information processing systems, 33:12816–12826, 2020

    Ulysse Marteau-Ferey, Francis Bach, and Alessandro Rudi. Non-parametric models for non- negative functions.Advances in neural information processing systems, 33:12816–12826, 2020

    2020

  10. [18]

    Second order conditions to decom- pose smooth functions as sums of squares.SIAM Journal on Optimization, 34(1):616–641, 2024

    Ulysse Marteau-Ferey, Francis Bach, and Alessandro Rudi. Second order conditions to decom- pose smooth functions as sums of squares.SIAM Journal on Optimization, 34(1):616–641, 2024

    2024

  11. [19]

    Jared S. Murray. Multiple imputation: A review of practical and theoretical findings.Statistical Science, 33(2):142–159, 2018

    2018

  12. [20]

    Missing data imputation using optimal transport

    Boris Muzellec, Julie Josse, Claire Boyer, and Marco Cuturi. Missing data imputation using optimal transport. InInternational Conference on Machine Learning, pages 7130–7140. PMLR, 2020

    2020

  13. [21]

    SIAM, 1994

    Yurii Nesterov and Arkadii Nemirovskii.Interior-point polynomial algorithms in convex programming. SIAM, 1994

    1994

  14. [22]

    What is a good imputation under mar missingness? arXiv preprint arXiv:2403.19196, 2024

    Jeffrey Näf, Erwan Scornet, and Julie Josse. What is a good imputation under mar missingness? arXiv preprint arXiv:2403.19196, 2024

    2024

  15. [23]

    Imputation scores.The Annals of Applied Statistics, 17(3):2452–2472, 2023

    Jeffrey Näf, Meta-Lina Spohn, Loris Michel, and Nicolai Meinshausen. Imputation scores.The Annals of Applied Statistics, 17(3):2452–2472, 2023

    2023

  16. [24]

    Multivariate beta distributions and independence properties of the wishart distribution.The Annals of Mathematical Statistics, pages 261–269, 1964

    Ingram Olkin and Herman Rubin. Multivariate beta distributions and independence properties of the wishart distribution.The Annals of Mathematical Statistics, pages 261–269, 1964

    1964

  17. [25]

    Scikit- learn: Machine learning in python.the Journal of machine Learning research, 12:2825–2830, 2011

    Fabian Pedregosa, Gaël Varoquaux, Alexandre Gramfort, Vincent Michel, Bertrand Thirion, Olivier Grisel, Mathieu Blondel, Peter Prettenhofer, Ron Weiss, Vincent Dubourg, et al. Scikit- learn: Machine learning in python.the Journal of machine Learning research, 12:2825–2830, 2011. 10

    2011

  18. [26]

    Psd representations for effective probability models

    Alessandro Rudi and Carlo Ciliberto. Psd representations for effective probability models. Advances in Neural Information Processing Systems, 34:19411–19422, 2021

    2021

  19. [27]

    Finding global minima via kernel approximations.arXiv preprint arXiv:2012.11978, 2020

    Alessandro Rudi, Ulysse Marteau-Ferey, and Francis Bach. Finding global minima via kernel approximations.arXiv preprint arXiv:2012.11978, 2020

    2012 arXiv

  20. [28]

    Cambridge University Press, 2014

    Shai Shalev-Shwartz and Shai Ben-David.Understanding Machine Learning: From Theory to Algorithms. Cambridge University Press, 2014

    2014

  21. [29]

    Sparse gaussian processes using pseudo-inputs

    Edward Snelson and Zoubin Ghahramani. Sparse gaussian processes using pseudo-inputs. Advances in neural information processing systems, 18, 2005

    2005

  22. [30]

    Missforest—non-parametric missing value imputation for mixed-type data.Bioinformatics, 28(1):112–118, 2012

    Daniel J Stekhoven and Peter Bühlmann. Missforest—non-parametric missing value imputation for mixed-type data.Bioinformatics, 28(1):112–118, 2012

    2012

  23. [31]

    The energy of data.Annual Review of Statistics and Its Application, 4(1):447–479, 2017

    Gábor J Székely and Maria L Rizzo. The energy of data.Annual Review of Statistics and Its Application, 4(1):447–479, 2017

    2017

  24. [32]

    Csdi: Conditional score-based diffusion models for probabilistic time series imputation.Advances in Neural Information Processing Systems, 34:24804–24816, 2021

    Yusuke Tashiro, Jiaming Song, Yang Song, and Stefano Ermon. Csdi: Conditional score-based diffusion models for probabilistic time series imputation.Advances in Neural Information Processing Systems, 34:24804–24816, 2021

    2021

  25. [33]

    Tsybakov.Introduction to Nonparametric Estimation

    Alexander B. Tsybakov.Introduction to Nonparametric Estimation. Springer, 2009

    2009

  26. [34]

    mice: Multivariate imputation by chained equations in r.Journal of Statistical Software, 45(3):1–67, 2011

    Stef Van Buuren and Karin Groothuis-Oudshoorn. mice: Multivariate imputation by chained equations in r.Journal of Statistical Software, 45(3):1–67, 2011

    2011

  27. [35]

    Springer, 2009

    Cédric Villani.Optimal Transport: Old and New. Springer, 2009

    2009

  28. [36]

    Are deep learning models superior for missing data imputation in large surveys? evidence from an empirical comparison

    Zhenhua Wang, Olanrewaju Akande, Jason Poulos, and Fan Li. Are deep learning models superior for missing data imputation in large surveys? evidence from an empirical comparison. arXiv preprint arXiv:2103.09316, 2021

    2021 arXiv

  29. [37]

    Numerical optimization, 2006

    Stephen J Wright. Numerical optimization, 2006

    2006

  30. [38]

    Gain: Missing data imputation using generative adversarial nets

    Jinsung Yoon, James Jordon, and Mihaela van der Schaar. Gain: Missing data imputation using generative adversarial nets. InInternational Conference on Machine Learning, pages 5689–5698. PMLR, 2018. 11 Appendix of Distributionally Faithful Imputation via Positive Semi–Definite ...

    2018

  31. [39]

    Our methodPSD_Imputeappears as a conditional mean estimator psd_mean and as a multiple imputer psd_multiple with m= 10 draws

    and consists of iterative imputation by conditional mean expectation.SoftImpute[ 9] is a low-rank method that alternates between solving least-squares problems and applying singular value thresholding.OT[ 20] is an optimal transport based method, which minimizes a suitable met...

  32. [40]

    2.fort= 0,1,

    InitialiseQ←I/ℓso thatTr(QH) = 1. 2.fort= 0,1, . . .do (a) Compute gradientand Hessian–vector products (b) Solve for Newton direction∆with SMW. (c) Newton decrementλ 2 =−⟨g(Q),∆⟩.ifλ 2/2≤εstop. (d) Backtracking line search: start α= 1 , reduce α←βα until f(Q+α∆) satisfies the ...

Pith tools

Reviewed July 10, 2026 · model on record in the stance chip above.