Pith. sign in

REVIEW 3 major objections 4 minor 1 cited by

Lecture Notes: Selected topics on robust statistical learning theory

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read These notes argue that four principles make robust estimators match Gaussian benchmarks.

desk verdict Useful synthesis with a real hole in the convex-loss homogeneity lemma; the unified framework doesn't hold together as proved, but the underlying literature is sound and the notes are readable. read the letter →

arxiv 1908.10761 v1 pith:D3QUNFY6 submitted 2019-08-28 stat.ML cs.LGmath.STstat.TH

Matthieu Lerasle This is my paper · ORCID
classification stat.MLcs.LGmath.STstat.TH MSC 62G3562G0562J05
keywords robuststatisticallearningmedian-of-meansminmaxestimatorshomogeneitylemmasmall-ballmethodsub-Gaussiandeviationsoracleinequalitiesleast-squaresregression
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

These lecture notes work toward a unified understanding of robust statistical learning. Their stated goal is to show that four principles\u2014median-of-means, the minmax aggregation of pairwise tests, the homogeneity lemma, and the small-ball method\u2014are enough to build estimators that obey the same oracle inequalities as ordinary empirical risk minimization does under sub-Gaussian data, while assuming only finite moments. If this is right, robustness is not a collection of unrelated tricks: the same template covers univariate mean estimation, Lipschitz-convex classification, least-squares regression, and Hellinger density estimation. The notes therefore supply a route from Gaussian benchmarks to heavy-tailed guarantees without redoing the argument case by case.

What carries the argument

The machinery has four named components. Median-of-means splits data into blocks and takes the median of block averages, turning a second-moment assumption into a level-dependent sub-Gaussian deviation bound. The minmax principle builds an estimator of the oracle as $\arg\min_f \sup_g \widehat{P}[\ell_f-\ell_g]$, using robust estimates of pairwise loss increments rather than of the loss itself. The homogeneity lemma (Lemma 60) reduces the risk analysis of such minmax estimators to deviation bounds of the test process on localized classes with level $E(f)\le r$, replacing peeling arguments when deviation bounds are available only up to a confidence level. The small-ball method supplies those localized deviation bounds for median-of-means processes under weak moment assumptions.

What would settle it

Take $d$ disjoint cells of equal probability, let $X$ be the vector of cell indicators, and choose $f$ supported on one cell; then $P[|X^T f|]/\sqrt{P[(X^T f)^2]}=1/\sqrt{d}$, so no absolute $\gamma$ satisfies (7.15) as $d$ grows. Checking the least-squares rates in Theorem 91 on this design\u2014or computing $C_Q(F)$ from (7.23) for the same design\u2014would show whether the small-ball route or the alternative localized analysis is the one that carries the argument.

Watch

Extended reading notes

Core claim

The central claim, stated in Section 1.3, is that the combination of these principles proves oracle inequalities simultaneously for the ERM in the sub-Gaussian framework, providing the relevant benchmarks, and for robust alternatives such as minmax MOM estimators. The paper's own examples certify the claim for univariate means, multivariate means under Euclidean or other norms, Lipschitz-convex losses such as SVM and boosting, linear least squares under a small-ball condition, and Hellinger density estimation via $\rho$-estimators. The unifying object is the test process $T(f,g)$ estimating $P[\ell_f-\ell_g]$; replacing the empirical mean by a median-of-means process inside a minmax estimator preserves the oracle inequality while weakening distributional assumptions.

Load-bearing premise

The load-bearing premise is the small-ball hypothesis (7.15): a single absolute constant $\gamma>0$ must satisfy $P[|X^T f|]\ge \gamma \sqrt{P[(X^T f)^2]}$ for every function in the model; for histogram-like designs this constant is $1/\sqrt{d}$, so the uniform version fails exactly when the design is spread across many localized cells.

Editorial extensions

If this is right

  • Mean estimation: MOM and Catoni-type estimators give $\sqrt{1/N}$ sub-Gaussian deviations with only two finite moments; the notes show no level-free sub-Gaussian estimator can exist over all distributions with two moments.
  • Lipschitz-convex losses: minmax MOM versions of SVM and boosting reach rates controlled by Rademacher complexity under moment assumptions on the design, where the ERM analysis needed Gaussian design.
  • Least-squares regression under the small-ball hypothesis has minmax MOM rates of order $\sigma(\sqrt{d}\vee\sqrt{K})/\sqrt{N}$ with only second moments; for histogram designs the small-ball constant degrades and an alternative complexity $C_Q(F)$ recovers optimal rates.
  • Density estimation: $\rho$-estimators, built from the same minmax principle and analysed with the homogeneity lemma, yield Hellinger oracle inequalities without assumptions on the target density or the model.
  • The median step confers resistance to a small proportion of arbitrary outliers in the $O\cup I$ model, with rates degrading by terms involving the outlier proportion.

Reading between the lines

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

  • Editorial extension: if the unified template is correct, robust estimation can be viewed as a compiler that replaces the linear empirical mean inside ERM by any sub-Gaussian univariate estimator; each new robust univariate construction automatically upgrades every minmax problem to which the homogeneity lemma applies.
  • Editorial extension: the histogram counterexample suggests that for designs built on localized basis functions the operative uniform parameter is dimension-dependent; a testable program is to compute $\gamma$ or $C_Q(F)$ explicitly for sparse high-dimensional dictionaries and see where the $\sqrt{d}\vee\sqrt{K}$ rates survive.
  • Editorial extension: the same analysis could be turned into a concrete experiment\u2014fit minmax MOM least squares and ERM on heavy-tailed histogram data with $d\approx\sqrt{N}$, and verify whether the MOM estimator keeps the predicted rate while ERM's confidence degrades.
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

3 major / 4 minor

Summary. These lecture notes aim to extract and present four principles of robust statistical learning theory: median-of-means estimators, minmax aggregation of tests, the homogeneity lemma, and the small-ball method. The central claim, stated in Section 1.3, is that the combination of these principles yields oracle inequalities simultaneously for the ERM in the sub-Gaussian framework and for robust alternatives such as minmax MOM estimators under weak moment assumptions. The notes develop the tools in Chapters 2-5 and then apply them to univariate and multivariate mean estimation, learning from Lipschitz-convex losses, least-squares regression, and density estimation with Hellinger loss. A short final chapter discusses polynomial-time computable estimators. The manuscript is explicitly a set of lecture notes and repeatedly acknowledges its own limitations, including the small-ball failure for histogram designs in Section 7.4 and the brevity of the computational chapter.

Significance. If the derivations were correct, this would be a valuable pedagogical synthesis: it connects several strands of the robust-statistics literature and organizes them around reusable principles, while benchmarking against classical results such as Hanson-Wright, Bousquet's inequality, and Berry-Esseen bounds. The notes are also honest about scope and do not fit parameters to data. However, the central homogeneity machinery for convex losses contains a sign-error gap that affects the proof of most subsequent oracle inequalities. The affected theorem statements may well be true and are drawn from the literature, but the notes as written do not prove them.

major comments (3)
  1. [§5.2.3, Lemma 63] The proof of the homogeneity property is invalid. Convexity gives T(f,f*) ≥ α T(fr,f*) with α = E(f)/r > 1, and the displayed conclusion T(f,f*) ≥ T(fr,f*) follows only if T(fr,f*) ≥ 0. Nothing in the assumptions guarantees this for empirical means or MOM operators, whose estimated increments can be negative even when the population risk of fr is larger than that of f*. The failure is not hypothetical: take F = R, f* = 0, squared loss, empirical mean with N = 1 and z = 10, r = 1, f = 2; then fr = 1, T(f,f*) = −36, and T(fr,f*) = −19, so T(f,f*) ≥ T(fr,f*) fails while the weaker α-inequality holds. Since α = E(f)/r is unbounded as E(f) grows, the localization argument in Lemma 60 cannot absorb the missing factor.
  2. [§5.2.2, Lemma 60 and Chapters 5-7] Because Lemma 60 uses the exact homogeneity property to pass from localized bounds to global risk bounds, the gap in Lemma 63 undermines the proofs of the theorems that invoke it: Theorem 65, Theorem 66, Theorem 82, Theorem 84, Theorem 91, and the homogeneity-lemma route to Corollary 95. In particular, for r = r1 in Lemma 60 the proof requires T(f*,f) ≤ T(f*,fr) for all f with E(f) > r1; the α-version only gives T(f*,f) ≤ α T(f*,fr), which is useless when T(f*,fr) is positive. The manuscript therefore needs either a corrected convex-loss homogeneity lemma or a modified homogeneity lemma that handles the multiplicative factor explicitly.
  3. [§5.2.4, Lemma 64] The same factor-sign issue appears in the proof that ρ-tests satisfy the homogeneity property. The derivation gives T(f,f*) ≥ (1/(1−ε)) T(fε,f*), and the conclusion T(f,f*) ≥ T(fε,f*) requires T(fε,f*) ≥ 0. This is not guaranteed for the ρ-statistic, which is a sum of values in [−1,1] and can be negative. Consequently, the analysis of ρ-estimators in Chapter 8, in particular the use of Lemma 60 in the proof of Theorem 101, inherits the same gap unless an additional argument is supplied.
minor comments (4)
  1. [§3.2.2, Lemma 35] The text states 'The proof of the lemma is omitted' for a lemma that is used in the proof of Bousquet's inequality. In a self-contained set of lecture notes this should either include a proof or give a precise reference for the calculus lemma.
  2. [§3.4, Theorem 38] Theorem 38 is described as having a proof that 'follows exactly the same arguments and is left to the reader.' Since this general concentration bound is used repeatedly in later chapters, a full proof or a detailed reference would improve the exposition.
  3. [Various proofs] Several 'standard density arguments' are invoked without details, for example in the proofs of Theorems 28, 30, and 37. These are indeed standard, but for lecture notes it would be helpful to state the relevant approximation or cite a single source where these arguments are carried out.
  4. [Throughout] There are several typographical and grammatical issues, such as 'Lipshitz' instead of 'Lipschitz' and 'tolerates much outliers sin ce' in Section 4.6.2. These do not affect the mathematics but should be corrected in a revision.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the notes derive their concentration and homogeneity tools inside the text and use citations only as provenance or external benchmarks.

full rationale

The paper's derivation chain is self-contained for its main machinery. The MOM deviation bounds in Chapter 3 (Theorems 37, 38, and 40) are proved in the notes using bounded-difference concentration, symmetrization, and contraction arguments; they are not imported as black boxes. The homogeneity lemma (Lemma 60) is stated and proved in Section 5.2.2, and the convex-loss homogeneity property (Lemma 63) is proved in Section 5.2.3. The statement that the homogeneity lemma extends a deterministic argument in [18], and that Chapter 6 presents results proved in [18], are provenance remarks rather than load-bearing citations: the actual inequalities are established in the text. No estimator parameter is fitted to a subset of data and then presented as a prediction; all results are explicit high-probability deviation bounds. The small-ball hypothesis (7.15) is an explicitly stated assumption, and Section 7.4 openly documents an important case where it fails, then provides a different analysis via Corollary 95 and Theorem 96. Invocations of external results such as Lugosi-Mendelson, Minsker-Strawn, Catoni, and Bousquet are either proved in the notes or used as standard independent benchmarks, not as self-citations that carry the derivation. In particular, no equation is shown to reduce by construction to its own input, and no fitted parameter is renamed as a prediction. The alleged invalidity of the convex-loss homogeneity proof, if correct, would be a correctness gap rather than circularity. Overall, the paper does not exhibit self-definitional, fitted-input, or self-citation-load-bearing circularity.

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

The notes introduce no new entities or fitted parameters; all assumptions are classical statistical model assumptions inherited from the cited literature.

assumptions (4)
  • domain assumption A minimizer f* of the risk exists in F (Section 1.1)
    Defines the excess risk E(f) = P[L_f - L_f*]; the author notes the results hold if inf replaces the minimizer.
  • domain assumption Small ball hypothesis: exists gamma > 0 such that for all f in F, P[|X^T f|] >= gamma sqrt(P[(X^T f)^2]) (Eq. 7.15)
    Key to the quadratic process bound in Theorem 91 for least-squares regression; fails for histogram designs (Section 7.4), requiring a separate analysis.
  • domain assumption Bernstein condition: exists A,B > 0 such that for all f with E(f) <= A, P[L_f - L_f*] >= B E(f)^2 (Eq. 6.15)
    Used to obtain fast rates for Lipschitz-convex losses in Theorems 82 and 84; checked for Huber and logistic losses in Propositions 79 and 80.
  • domain assumption L4/L2 comparison: exists Delta >= 1 such that for all f in F, ||f - f*||_{L4} <= Delta ||f - f*||_{L2} (Eq. 6.17)
    Used in the Bernstein condition proofs and in Section 6.6; Proposition 77 shows it may degrade with dimension for histogram spaces.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Lecture Notes: Selected topics on robust statistical learning theory." pith.science (2026). https://pith.science/paper/D3QUNFY6

@misc{pith2026190810761,
  author       = {Pith},
  title        = {Pith review of: Lecture Notes: Selected topics on robust statistical learning theory},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D3QUNFY6}},
  note         = {Machine review of arXiv:1908.10761}
}
read the original abstract

These notes gather recent results on robust statistical learning theory. The goal is to stress the main principles underlying the construction and theoretical analysis of these estimators rather than provide an exhaustive account on this rapidly growing field. The notes are the basis of lectures given at the conference StatMathAppli 2019.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Open Problem: Is Interaction Necessary for Order-Optimal 1-bit Mean Estimation?

    cs.IT 2026-07 conditional novelty 5.0 of 10

    Is interaction necessary for order-optimal 1-bit mean estimation over nonparametric finite-moment classes, or can fully non-adaptive general quantizers match the adaptive rate?

Reference graph

Works this paper leans on

58 extracted references · 50 canonical work pages · cited by 1 Pith paper

  1. [1]

    Estimation b ounds and sharp oracle inequalities of regularized procedures with Lipschit z loss functions

    Pierre Alquier, Vincent Cottet, and Guillaume Lecu´ e. Estimation b ounds and sharp oracle inequalities of regularized procedures with Lipschit z loss functions. Ann. Statist. , 47(4):2117–2144, 2019

    work page 2019

  2. [2]

    Robust linear least square s regres- sion

    Jean-Yves Audibert and Olivier Catoni. Robust linear least square s regres- sion. Ann. Statist. , 39(5):2766–2794, 2011

    work page 2011

  3. [3]

    Baraud and L

    Y. Baraud and L. Birg´ e. Rho-estimators for shape restricted density esti- mation. Stochastic Process. Appl. , 126(12):3888–3912, 2016

    work page 2016

  4. [4]

    Baraud, L

    Y. Baraud, L. Birg´ e, and M. Sart. A new method for estimation a nd model selection: ρ-estimation. Invent. Math. , 207(2):425–517, 2017

    work page 2017

  5. [5]

    Non-asymptotic minimax rates of testing in signa l detec- tion

    Yannick Baraud. Non-asymptotic minimax rates of testing in signa l detec- tion. Bernoulli, 8(5):577–606, 2002

    work page 2002

  6. [6]

    Rho-estimators revisited: general theory and applications

    Yannick Baraud and Lucien Birg´ e. Rho-estimators revisited: general theory and applications. Ann. Statist. , 46(6B):3767–3804, 2018

    work page 2018

  7. [7]

    Stabilit´ e et instabilit´ e du risque minimax pour des variables ind´ ependantes ´ equidistribu´ ees.Ann

    Lucien Birg´ e. Stabilit´ e et instabilit´ e du risque minimax pour des variables ind´ ependantes ´ equidistribu´ ees.Ann. Inst. H. Poincar´ e Probab. Statist. , 20(3):201–223, 1984

    work page 1984

  8. [8]

    Model selection via testing: an alternative to (pen alized) maximum likelihood estimators

    Lucien Birg´ e. Model selection via testing: an alternative to (pen alized) maximum likelihood estimators. Ann. Inst. H. Poincar´ e Probab. Statist. , 42(3):273–325, 2006

    work page 2006

Show all 58 references
  1. [9]

    Rates of convergence for m inimum con- trast estimators

    Lucien Birg´ e and Pascal Massart. Rates of convergence for m inimum con- trast estimators. Probab. Theory Related Fields , 97(1-2):113–150, 1993

    1993

  2. [10]

    Concentration Inequalities: A Nonasymptotic Theory of Independence

    St´ ephane Boucheron, G´ abor Lugosi, and Pascal Massart. Concentration Inequalities: A Nonasymptotic Theory of Independence. Oxford University Press, 2013. ISBN 978-0-19-953525-5

    2013

  3. [11]

    A Bennett concentration inequality and its app lication to suprema of empirical processes

    Olivier Bousquet. A Bennett concentration inequality and its app lication to suprema of empirical processes. C. R. Math. Acad. Sci. Paris , 334(6):495– 500, 2002

    2002

  4. [12]

    Catoni and I

    O. Catoni and I. Giulini. Dimension-free pac-bayesian bounds fo r matri- ces, vectors, and linear least squares regression. Technical rep ort, 2017. https://arxiv.org/pdf/1712.02747.pdf. 139 140 BIBLIOGRAPHY

    2017 arXiv

  5. [13]

    Statistical learning theory and stochastic optimization , vol- ume 1851 of Lecture Notes in Mathematics

    Olivier Catoni. Statistical learning theory and stochastic optimization , vol- ume 1851 of Lecture Notes in Mathematics . Springer-Verlag, Berlin, 2004. Lecture notes from the 31st Summer School on Probability Theory held in Saint-Flour, July 8–25, 2001

    2004

  6. [14]

    Challenging the empirical mean and empirical varianc e: a deviation study

    Olivier Catoni. Challenging the empirical mean and empirical varianc e: a deviation study. Ann. Inst. Henri Poincar´ e Probab. Stat., 48(4):1148–1185, 2012

    2012

  7. [15]

    Robust covariance and scat- ter matrix estimation under Huber’s contamination model

    Mengjie Chen, Chao Gao, and Zhao Ren. Robust covariance and scat- ter matrix estimation under Huber’s contamination model. Ann. Statist. , 46(5):1932–1960, 2018

    1932

  8. [16]

    High-dimensional robu st mean estimation in nearly-linear time

    Yu Cheng, Ilias Diakonikolas, and Rong Ge. High-dimensional robu st mean estimation in nearly-linear time. In Proceedings of the Thirtieth Annual ACM-SIAM Symposium on Discrete Algorithms , SODA ’19, pages 2755– 2771, Philadelphia, PA, USA, 2019. Society for Industrial and App...

    2019

  9. [17]

    B artlett

    Yeshwanth Cherapanamjeri, Nicolas Flammarion, and Peter L. B artlett. Fast mean estimation with sub-gaussian rates. In Alina Beygelzimer a nd Daniel Hsu, editors, Proceedings of the Thirty-Second Conference on Learn- ing Theory, volume 99 of Proceedings of Machine Learning Res...

    2019

  10. [18]

    Chinot, G

    G. Chinot, G. Lecu´ e, and M. Lerasle. Statistical learning with lip schitz and convex loss functions. to appear in Probab. Theory Related Fields, arXiv:1810.01090, 2019

    2019 arXiv

  11. [19]

    Robust subgaussian estima tion of a mean vector in nearly linear time

    Jules Depersin and Guillaume Lecu´ e. Robust subgaussian estima tion of a mean vector in nearly linear time. 1906.03058, 2019

    1906 arXiv

  12. [20]

    Oliveira

    Luc Devroye, Matthieu Lerasle, Gabor Lugosi, and Roberto I. Oliveira. Sub-Gaussian mean estimators. Ann. Statist. , 44(6):2695–2725, 2016

    2016

  13. [21]

    Family- wise separation rates for multiple testing

    Magalie Fromont, Matthieu Lerasle, and Patricia Reynaud-Bour et. Family- wise separation rates for multiple testing. Ann. Statist. , 44(6):2533–2563, 2016

    2016

  14. [22]

    Contribution to the theory of robust estimatio n

    Frank R Hampel. Contribution to the theory of robust estimatio n. Ph. D. Thesis, University of California, Berkeley , 1968

    1968

  15. [23]

    A general qualitative definition of robustness

    Frank R Hampel. A general qualitative definition of robustness. The Annals of Mathematical Statistics , pages 1887–1896, 1971

    1971

  16. [24]

    Robust estimation: A condensed partial surve y

    Frank R Hampel. Robust estimation: A condensed partial surve y. Proba- bility Theory and Related Fields , 27(2):87–104, 1973

    1973

  17. [25]

    The influence curve and its role in robust estimat ion

    Frank R Hampel. The influence curve and its role in robust estimat ion. Journal of the American Statistical Association , 69(346):383–393, 1974

    1974

  18. [26]

    Beyond location parameters: Robust concept s and meth- ods

    Frank R Hampel. Beyond location parameters: Robust concept s and meth- ods. Bulletin of the International statistical Institute , 46(1):375–382, 1975. BIBLIOGRAPHY 141

    1975

  19. [27]

    Samuel B. Hopkins. Fast mean estimation with sub-gaussian rat es. to appear in Ann. Statist., arXiv:1809.07425 , 2018

    2018 arXiv

  20. [28]

    The behavior of maximum likelihood estimates unde r non- standard conditions

    Peter J Huber. The behavior of maximum likelihood estimates unde r non- standard conditions. In Proceedings of the fifth Berkeley symposium on mathematical statistics and probability , volume 1, pages 221–233. Berkeley, CA, 1967

    1967

  21. [29]

    Robust estimation of a location parameter

    Peter J Huber et al. Robust estimation of a location parameter. The Annals of Mathematical Statistics , 35(1):73–101, 1964

    1964

  22. [30]

    Huber and Elvezio M

    Peter J. Huber and Elvezio M. Ronchetti. Robust statistics. Wiley Series in Probability and Statistics. John Wiley & Sons, Inc., Hoboken, NJ, sec ond edition, 2009

    2009

  23. [31]

    Oracle inequalities in empirical risk minimization and sparse recovery problems , volume 2033 of Lecture Notes in Mathe- matics

    Vladimir Koltchinskii. Oracle inequalities in empirical risk minimization and sparse recovery problems , volume 2033 of Lecture Notes in Mathe- matics. Springer, Heidelberg, 2011. Lectures from the 38th Probability Summer School held in Saint-Flour, 2008, ´Ecole d’´Et´ e de Pr...

    2011

  24. [32]

    Bounding the smallest sin- gular value of a random matrix without concentration

    Vladimir Koltchinskii and Shahar Mendelson. Bounding the smallest sin- gular value of a random matrix without concentration. Int. Math. Res. Not. IMRN , (23):12991–13008, 2015

    2015

  25. [33]

    Lecu´ e and M

    G. Lecu´ e and M. Lerasle. Learning from mom’s principle : Le cam’s ap- proach. Technical report, CNRS, ENSAE, Paris-sud. To appear in Stoch. Proc. App

  26. [34]

    Lecu´ e and M

    G. Lecu´ e and M. Lerasle. Robust machine learning by median-of -means : theory and practice. to appear in Ann. Statist., arXiv:1711.10306 , 2019

    2019 arXiv

  27. [35]

    Lecu´ e, M

    G. Lecu´ e, M. Lerasle, and T. Mathieu. Robust classification via mom minimization. arXiv preprint arXiv:1808.03106 , 2018

    2018 arXiv

  28. [36]

    The concentration of measure phenomenon , volume 89 of Mathematical Surveys and Monographs

    Michel Ledoux. The concentration of measure phenomenon , volume 89 of Mathematical Surveys and Monographs . American Mathematical Society, Providence, RI, 2001

    2001

  29. [37]

    Lei, K Luh, P Venkat, and F

    Z. Lei, K Luh, P Venkat, and F. Zhang. A fast spectral algorith m for mean estimation withsub-gaussian rates. Arxive:1908.04468, 2019

    1908 arXiv

  30. [38]

    Lugosi and S

    G. Lugosi and S. Mendelson. Robust multivariate mean estimatio n: the optimality of trimmed mean. 1907.11391, 2019

    1907 arXiv

  31. [39]

    Risk minimization by median- of- means tournaments

    Gabor Lugosi and Shahar Mendelson. Risk minimization by median- of- means tournaments. To appear in JEMS

  32. [40]

    Mean estimation and reg ression under heavy-tailed distributions: A survey

    G´ abor Lugosi and Shahar Mendelson. Mean estimation and reg ression under heavy-tailed distributions: A survey. Foundations of Computational Mathematics, Aug 2019

    2019

  33. [41]

    Regularization, sparse recovery, and median-of-means tournaments

    G´ abor Lugosi and Shahar Mendelson. Regularization, sparse recovery, and median-of-means tournaments. Bernoulli, 25(3):2075–2106, 2019. 142 BIBLIOGRAPHY

    2019

  34. [42]

    Sub-Gaussian estimators of the mean of a random vector

    G´ abor Lugosi and Shahar Mendelson. Sub-Gaussian estimators of the mean of a random vector. Ann. Statist. , 47(2):783–794, 2019

    2019

  35. [43]

    Some applications of concentration inequalitie s to statis- tics

    Pascal Massart. Some applications of concentration inequalitie s to statis- tics. Ann. Fac. Sci. Toulouse Math. (6) , 9(2):245–303, 2000. Probability theory

    2000

  36. [44]

    Risk bounds for statistical learning

    Pascal Massart and ´Elodie N´ ed´ elec. Risk bounds for statistical learning. Ann. Statist. , 34(5):2326–2366, 2006

    2006

  37. [45]

    Learning without concentration

    Shahar Mendelson. Learning without concentration. In Proceedings of the 27th annual conference on Learning Theory COLT14, pages pp 25–39. 2014

    2014

  38. [46]

    Minsker and N

    S. Minsker and N. Strawn. Distributed statistical estimation an d rates of convergence in normal approximation. Preprint available on arXiv:1704.02658

    arXiv

  39. [47]

    Geometric median and robust estimation in Ban ach spaces

    Stanislav Minsker. Geometric median and robust estimation in Ban ach spaces. Bernoulli, 21(4):2308–2335, 2015

    2015

  40. [48]

    A. S. Nemirovsky and D. B. Yudin. Problem complexity and method effi- ciency in optimization . A Wiley-Interscience Publication. John Wiley & Sons, Inc., New York, 1983. Translated from the Russian and with a preface by E. R. Dawson, Wiley-Interscience Series in Discrete Mathematics

    1983

  41. [49]

    Faster and simpler width - independent parallel algorithms for positive semidefinite programmin g

    Richard Peng and Kanat Tangwongsan. Faster and simpler width - independent parallel algorithms for positive semidefinite programmin g. In Proceedings of the Twenty-fourth Annual ACM Symposium on Pa rallelism in Algorithms and Architectures , SPAA ’12, pages 101–108, New York, N...

    2012

  42. [50]

    Prasad, A.S

    A. Prasad, A.S. Suggala, S. Balakrishnan, and P. Ravikumar. Ro bust estimation via robust gradient estimation. Arxive:1802.06485, 2018

    2018 arXiv

  43. [51]

    On optimality of empirical risk minimization in linear aggregation

    Adrien Saumard. On optimality of empirical risk minimization in linear aggregation. Bernoulli, 24(3):2176–2203, 2018

    2018

  44. [52]

    Support Vector Machines

    Ingo Steinwart and Andreas Christmann. Support Vector Machines . Springer, 2008

    2008

  45. [53]

    Tsybakov

    Alexandre B. Tsybakov. Optimal aggregation of classifiers in st atistical learning. Ann. Statist. , 32(1):135–166, 2004

    2004

  46. [54]

    A survey of sampling from contaminated distribut ions

    John W Tukey. A survey of sampling from contaminated distribut ions. Contributions to probability and statistics , 2:448–485, 1960

    1960

  47. [55]

    The future of data analysis

    John W Tukey. The future of data analysis. The annals of mathematical statistics, 33(1):1–67, 1962

    1962

  48. [56]

    J.W. Tukey. Adress to international congress of mathematicia ns. Technical report, Vancouver, 1974

    1974

  49. [57]

    J.W. Tukey. T6: Order statistics. Technical report, In mimeog raphed notes for Statistics 411, Princeton Univ., 1974. BIBLIOGRAPHY 143

    1974

  50. [58]

    Vladimir N. Vapnik. The nature of statistical learning theory . Statistics for Engineering and Information Science. Springer-Verlag, New York, second edition, 2000

    2000

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.