pith. sign in

arxiv: 2603.29739 · v2 · submitted 2026-03-31 · 🧮 math.PR

From oracle maximal inequalities to martingale random fields via finite approximation from below

Yoichi Nishiyama This is my paper

Pith reviewed 2026-05-13 23:29 UTC · model grok-4.3

classification 🧮 math.PR
keywords martingale random fieldsmaximal inequalitiesLenglart inequalityweak convergenceDonsker propertyempirical processesfinite approximationsubmartingales
0
0 comments X

The pith

A novel oracle maximal inequality via integration by parts yields sharp bounds for martingale random field suprema.

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

This paper proposes a new method to derive sharp upper bounds on the expected supremum of separable martingale random fields. It starts with an oracle maximal inequality for finite classes of submartingales derived using integration by parts rather than the triangle inequality. This leads to a generalization of Lenglart's inequality to finite-dimensional cases and further to infinite-dimensional ones through finite approximation from below. The results provide new weak convergence theorems and conditions for the Donsker property in empirical processes.

Core claim

The paper establishes that an oracle maximal inequality for a finite class of submartingales, obtained through integration by parts, can be used to bound the supremum of separable martingale random fields. By means of a finite approximation device, this approach extends Lenglart's inequality from one dimension to higher dimensions and certain infinite-dimensional settings, facilitating applications in weak convergence under uniform topology.

What carries the argument

The oracle maximal inequality for finite submartingales derived by integration by parts, together with the finite approximation device from below.

If this is right

  • Generalizes Lenglart's inequality to finite-dimensional martingales.
  • Extends the inequality to certain infinite-dimensional martingale random fields.
  • Provides new weak convergence theorems for martingale random fields.
  • Establishes a necessary and sufficient condition for the Donsker property of countable function classes.
  • Delivers new moment bounds for suprema of empirical processes.

Where Pith is reading between the lines

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

  • The approach could reduce reliance on chaining methods in stochastic process theory.
  • Finite approximations might enable computational verification of the bounds in practice.
  • Similar techniques may apply to other maximal inequalities in probability.
  • Separability ensures the approximation works without losing sharpness.

Load-bearing premise

The finite approximation device from below extends the finite-dimensional bound to infinite-dimensional separable martingale random fields without losing sharpness.

What would settle it

Finding a separable martingale random field where the expected supremum exceeds the bound given by the oracle maximal inequality would disprove the sharpness claim.

read the original abstract

A novel approach is proposed to establish a sharp upper bound on the expected supremum of a separable martingale random field, serving as an alternative to classical universal chaining-based methods. The proposed approach begins by deriving a new "oracle maximal inequality" for a finite class of submartingales. This is achieved via integration by parts rather than a simplistic application of the triangle inequality. Consequently, we obtain a generalization of Lenglart's inequality for discrete-time martingales, extending it from the one-dimensional case to finite-dimensional settings, and further to certain infinite-dimensional cases through a "finite approximation device". The primary applications include several weak convergence theorems for sequences of separable martingale random fields under the uniform topology. In particular, new results are established for i.i.d. sequences, including a necessary and sufficient condition for a countable class of functions to possess the Donsker property. Additionally, we provide new moment bounds for the supremum of empirical processes indexed by classes of sets or functions.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The paper proposes a novel approach to derive sharp upper bounds on the expected supremum of separable martingale random fields. It begins by establishing an oracle maximal inequality for finite classes of submartingales via integration by parts (rather than triangle inequality), yielding a generalization of Lenglart's inequality to finite-dimensional and certain infinite-dimensional settings through a finite approximation device from below. The method is then applied to obtain weak convergence theorems under the uniform topology, including a necessary and sufficient condition for the Donsker property of countable classes and new moment bounds for suprema of empirical processes.

Significance. If the finite approximation device preserves sharpness without introducing dimension-dependent gaps, the work supplies a useful alternative to classical chaining arguments for maximal inequalities on martingale random fields. This could streamline proofs of uniform convergence results and furnish new characterizations of Donsker classes, particularly when the index set is countable or the processes admit a separable version.

major comments (2)
  1. [finite approximation device] The finite approximation device (described after the oracle inequality for finite classes) must be shown to transmit the sharp constant to the separable infinite-dimensional case. The manuscript needs to supply an explicit error bound demonstrating that the approximating finite submartingales capture the essential supremum without a multiplicative factor that grows with the index set or dimension; separability alone does not automatically guarantee this, as noted in the stress-test concern.
  2. [oracle maximal inequality] The integration-by-parts step that produces the oracle maximal inequality for finite classes should be checked for hidden constants or assumptions that might affect sharpness when the class size increases; the abstract claims the bound is obtained without simplistic triangle inequality, but the precise dependence on the finite cardinality must be tracked through to the infinite-dimensional limit.
minor comments (1)
  1. [Abstract] The abstract would benefit from a brief statement of the error control or sharpness verification for the approximation step, as the current outline leaves the passage from finite to infinite dimensions somewhat opaque.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments, which have helped us identify points where the manuscript can be clarified. We address each major comment below and will revise the paper to incorporate explicit bounds and remarks on constant independence.

read point-by-point responses

  1. Referee: The finite approximation device (described after the oracle inequality for finite classes) must be shown to transmit the sharp constant to the separable infinite-dimensional case. The manuscript needs to supply an explicit error bound demonstrating that the approximating finite submartingales capture the essential supremum without a multiplicative factor that grows with the index set or dimension; separability alone does not automatically guarantee this, as noted in the stress-test concern.

    Authors: We agree that an explicit transmission result strengthens the argument. In the revision we will add a new lemma (Lemma 3.4) immediately after the oracle inequality, proving that for any separable martingale random field the finite approximations from below satisfy E[sup X] = lim E[sup X^n] with the same constant as in the finite-class oracle bound and with no multiplicative factor depending on cardinality or dimension. The proof uses the separability assumption to extract a countable dense index set whose finite subsets approximate the essential supremum monotonically from below; the error term vanishes by monotone convergence without introducing dimension-dependent gaps. We will also include a brief stress-test example confirming sharpness is preserved. revision: yes

  2. Referee: The integration-by-parts step that produces the oracle maximal inequality for finite classes should be checked for hidden constants or assumptions that might affect sharpness when the class size increases; the abstract claims the bound is obtained without simplistic triangle inequality, but the precise dependence on the finite cardinality must be tracked through to the infinite-dimensional limit.

    Authors: Re-examination of the integration-by-parts derivation confirms that the resulting constant is absolute and independent of the finite cardinality m. The formula is applied directly to the joint submartingale (max_{i=1..m} X_i, sum_{i=1..m} dX_i) and yields a bound whose leading constant does not grow with m; the avoidance of the triangle inequality is precisely what eliminates any m-dependent factor. This independence carries verbatim through the finite-approximation limit. We will insert a short remark after the statement of the oracle inequality explicitly recording that the constant is universal in m and remains unchanged in the passage to the separable infinite-dimensional case. revision: partial

Circularity Check

0 steps flagged

No circularity: derivation proceeds from integration by parts on finite classes to approximation device without self-referential reduction

full rationale

The paper's central chain starts with an oracle maximal inequality obtained via integration by parts for a finite class of submartingales, then invokes a finite approximation device to pass to separable martingale random fields. This construction is self-contained and does not define the target bound in terms of itself, rename a fitted quantity as a prediction, or rest on a load-bearing self-citation whose content reduces to the present result. The extension to infinite dimensions is presented as a mathematical device rather than a tautological renaming or ansatz smuggled via prior work. The derivation therefore retains independent content relative to its inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on classical martingale theory and the effectiveness of the newly introduced finite approximation device; no explicit free parameters or new physical entities are stated in the abstract.

axioms (1)
  • standard math Standard definitions and properties of discrete-time martingales and submartingales hold in finite and infinite dimensions
    The paper invokes these to derive the oracle inequality and its extensions.

pith-pipeline@v0.9.0 · 5463 in / 1324 out tokens · 46962 ms · 2026-05-13T23:29:51.735533+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

33 extracted references · 33 canonical work pages

  1. [1]

    and Levental, S

    Bae, J. and Levental, S. (1995). Uniform CLT for Markov chains and its invariance principle: a martingale approach.J. Theoret. Probab.8, 549-570

    work page 1995

  2. [2]

    and Tsybakov, A.B

    Bickel, P.J., Ritov, Y. and Tsybakov, A.B. (2009). Simultaneous analysis of LASSO and Dantzig selector.Ann. Statist.37, 1705-1732

    work page 2009

  3. [3]

    and van de Geer, S

    B˝ uhlmann, P. and van de Geer, S. (2011).Statistics for High-Dimensional Data. Springer-Verlag, New York

    work page 2011

  4. [4]

    and Tao, T

    Cand` es, E. and Tao, T. (2007). The Dantzig selector: Statistical estimation whenp is much larger thann.Ann. Statist.35, 3213-2351

    work page 2007

  5. [5]

    and Louhichi, S

    Dedecker, J. and Louhichi, S. (2002). Maximal inequalities and empirical central limit theorems.In: Empirical Process Techniques for Dependent Data.Edited by Dehling, H., Mikosch, T. and Sørensen, M. Birkh¨ auser, Boston

    work page 2002

  6. [6]

    (1953).Stochastic Processes.John Wiley & Sons, New York

    Doob, J.L. (1953).Stochastic Processes.John Wiley & Sons, New York

    work page 1953

  7. [7]

    and Rio, E

    Doukhan, P., Massart, P. and Rio, E. (1995). Invariance principles for absolutely regular empirical processes.Ann. Inst. Henri Poincar´ e Probab. Statist.31, 393-427

    work page 1995

  8. [8]

    Dudley, R.M. (1978). Central limit theorems for empirical measures.Ann. Probab.6, 899-929. Corrections,7, 909-911

    work page 1978

  9. [9]

    (1999).Uniform Central Limit Theorems.Cambridge University Press, Cambridge

    Dudley, R.M. (1999).Uniform Central Limit Theorems.Cambridge University Press, Cambridge

    work page 1999

  10. [10]

    and Nishiyama, Y

    Fujimori, K. and Nishiyama, Y. (2017a). Thel q consistency of the Dantzig selector for Cox’s proportional hazards model.J. Statist. Plann. Inference181, 62-70

    work page

  11. [11]

    and Nishiyama, Y

    Fujimori, K. and Nishiyama, Y. (2017b). The Dantzig selector for diffusion processes with covariates.J. Japan Statist. Soc.4759-73

    work page

  12. [12]

    and Marcus, M

    Jain, N. and Marcus, M. (1975). Central limit theorem forC(S)-valued random variables.J. Funct. Anal.19, 216-231

    work page 1975

  13. [13]

    and Shiryaev, A.N

    Jacod, J. and Shiryaev, A.N. (2003).Limit Theorems for Stochastic Processes.(2nd ed.) Springer-Verlag, Berlin Heidelberg

    work page 2003

  14. [14]

    Kolˇ cinskii, V.I. (1981). On the central limit theorems for empirical measures.Theory Probability and Math. Statist.24, 71-82. 24

    work page 1981

  15. [15]

    and Talagrand, M

    Ledoux, M. and Talagrand, M. (1991).Probability in Banach Spaces.Springer- Verlag, Berlin Heidelberg

    work page 1991

  16. [16]

    Lenglart, E. (1977). Relation de domination entre deux processus.Ann. Inst. Henri Poincar´ e(B)13, 171-179

    work page 1977

  17. [17]

    Levental, S. (1989). A uniform CLT for uniformly bounded families of martingale differences.J. Theoret. Probab.2, 271-287

    work page 1989

  18. [18]

    Nishiyama, Y. (1997). Some central limit theorems forℓ ∞-valued semimartingales and their applications.Probab. Theory Relat. Fields108, 459-494

    work page 1997

  19. [19]

    Nishiyama, Y. (1999). A maximal inequality for continuous martingales andM- estimation in a Gaussian white noise model.Ann. Statist.27, 675-696

    work page 1999

  20. [20]

    Nishiyama, Y. (2000a). Weak convergence of some classes of martingales with jumps. Ann. Probab.28, 685-712

    work page

  21. [21]

    (2000b).Entropy Methods for Martingales.CWI Tract128, Centrum voor Wiskunde en Informatica, Amsterdam

    Nishiyama, Y. (2000b).Entropy Methods for Martingales.CWI Tract128, Centrum voor Wiskunde en Informatica, Amsterdam

    work page

  22. [22]

    Weak convergence of some classes of martin- gales with jumps

    Nishiyama, Y. (2007). On the paper “Weak convergence of some classes of martin- gales with jumps”.Ann. Probab.35, 1194-1200

    work page 2007

  23. [23]

    (2022).Martingale Methods in Statistics.Chapman & Hall/CRC, Boca Raton, London, New York, Oxford

    Nishiyama, Y. (2022).Martingale Methods in Statistics.Chapman & Hall/CRC, Boca Raton, London, New York, Oxford

    work page 2022

  24. [24]

    Ossiander, M. (1987). A central limit theorem under metric entropy withL 2- bracketing.Ann. Probab.15, 897-919

    work page 1987

  25. [25]

    Pollard, D. (1982). A central limit theorem for empirical processes.J. Austral. Math. Soc. Ser. A33, 235-248

    work page 1982

  26. [26]

    Prokhorov, Yu. (1956). Convergence of random processes and limit theorems in probability theory.Theory Probab. Appl.1, 157-214

    work page 1956

  27. [27]

    Sudakov, V.N. (1969). Gaussian measurs, Cauchy measures andε-entropy.Soviet Math. Dok.10, 310-313

    work page 1969

  28. [28]

    Talagrand, M. (1987). Regularity of Gaussian processes.Acta Mathematica159, 99-149

    work page 1987

  29. [29]

    (2021).Upper and Lower Bounds for Stochastic Processes.(2nd ed.) Springer-Verlag, Berlin Heidelberg

    Talagrand, M. (2021).Upper and Lower Bounds for Stochastic Processes.(2nd ed.) Springer-Verlag, Berlin Heidelberg

    work page 2021

  30. [30]

    Tibshirani, E. (1996). Regression shrinkage and selection via the lasso.J. Royal Statist. Soc. Ser. B36, 111-147

    work page 1996

  31. [31]

    (2000).Empirical Processes inM-Estimation.Cambridge University Press, Cambridge

    van de Geer, S. (2000).Empirical Processes inM-Estimation.Cambridge University Press, Cambridge

    work page 2000

  32. [32]

    and Wellner, J.A

    van der Vaart, A.W. and Wellner, J.A. (2023).Weak Convergence and Empirical Processes: With Applications to Statistics.(2nd ed.) Springer-Verlag, New York

    work page 2023

  33. [33]

    predictable L-domination

    van der Vaart, A.W. and van Zanten, H. (2005). Donsker theorems for diffusion: necessary and sufficient conditions.Ann. Probab.33, 1422-1451. 25 Appendix: Lenglart’s inequality (for predictable time) Let a discrete-time stochastic basis (Ω,F,(F n)n∈N0,P) be given. Definition 25(L-domination; predictableL-domination).Let a non-negative adapted processX= (X...

    work page 2005