pith. sign in

arxiv: 2408.06531 · v2 · submitted 2024-08-12 · 💱 q-fin.RM · math.PR· q-fin.CP

Adaptive Multilevel Stochastic Approximation of the Value-at-Risk

St\'ephane Cr\'epey , Noufel Frikha , Azar Louzi , Jonathan Spence This is my paper

Pith reviewed 2026-05-23 21:43 UTC · model grok-4.3

classification 💱 q-fin.RM math.PRq-fin.CP
keywords value-at-riskmultilevel Monte Carlostochastic approximationadaptive samplingcomplexity boundsfinancial risk
0
0 comments X

The pith

An adaptive multilevel stochastic approximation algorithm computes Value-at-Risk with complexity O(ε^{-2} |ln ε|^{5/2}).

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

The paper proposes a multilevel stochastic approximation scheme for computing the Value-at-Risk of a financial loss that is only simulatable by Monte Carlo. It introduces an adaptive selection of the number of inner samples at each level to handle the bias from the discontinuous Heaviside function in the stochastic gradient. This achieves a complexity of O(ε^{-2} |ln ε|^{5/2}), improving on the prior O(ε^{-5/2}) bound. A reader would care as it makes accurate risk estimation more computationally feasible.

Core claim

The central claim is that adaptively selecting the number of inner samples at each level in the multilevel stochastic approximation allows mitigating the suboptimality caused by the Heaviside discontinuity, resulting in the best complexity O(ε^{-2} |ln ε|^{5/2}) for the algorithm.

What carries the argument

The adaptive rule that selects the number of inner samples at each level to control bias in the recursive stochastic gradient evaluation.

If this is right

  • The algorithm provides a more efficient way to estimate Value-at-Risk using Monte Carlo simulations.
  • It reduces the exponent in the complexity from 5/2 to 2, with only a logarithmic factor.
  • Numerical experiments exemplify the theoretical analysis and confirm the improved performance.

Where Pith is reading between the lines

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

  • This adaptive strategy could be applied to other problems involving discontinuous functions in stochastic gradients.
  • Extensions might include combining it with variance reduction techniques for further gains.
  • Similar adaptations may benefit computation of other quantiles or risk measures.

Load-bearing premise

The loss random variable satisfies regularity conditions that permit the adaptive inner-sample selection to control the bias from the discontinuous Heaviside function.

What would settle it

A numerical experiment where the observed complexity does not match O(ε^{-2} |ln ε|^{5/2}) for a loss variable meeting the regularity conditions would falsify the result.

Figures

Figures reproduced from arXiv: 2408.06531 by Azar Louzi, Jonathan Spence, Noufel Frikha, St\'ephane Cr\'epey.

Figure 5.1
Figure 5.1. Figure 5.1: Performance comparison of the different SA schemes. 1 32 1 64 1 128 1 256 1 512 Prescribed accuracy 10 4 10 3 10 2 10 1 10 0 10 1 A v era g e e x e c u tio n tim e (s) Value-at-risk NSA -adNSA adNSA MLSA adMLSA -adMLSA SA [PITH_FULL_IMAGE:figures/full_fig_p025_5_1.png] view at source ↗
Figure 5.2
Figure 5.2. Figure 5.2: Complexity comparison of the different SA schemes. 25 [PITH_FULL_IMAGE:figures/full_fig_p025_5_2.png] view at source ↗
Figure 5.3
Figure 5.3. Figure 5.3: Performance comparison of the different SA schemes. 28 [PITH_FULL_IMAGE:figures/full_fig_p028_5_3.png] view at source ↗
Figure 5.4
Figure 5.4. Figure 5.4: Complexity comparison of the different SA schemes. SA scheme NSA σ-adNSA adNSA MLSA adMLSA σ-adMLSA SA RMSE −3.52 −4.10 −3.86 −3.93 −2.60 −2.75 −1.66 Accuracy ε −3.00 −2.90 −2.78 −2.90 −1.96 −2.05 −2.00 [PITH_FULL_IMAGE:figures/full_fig_p029_5_4.png] view at source ↗
read the original abstract

Cr\'epey, Frikha, and Louzi (2025) introduced a multilevel stochastic approximation scheme to compute the value-at-risk of a financial loss that is only simulatable by Monte Carlo. The best complexity of the scheme is in O($\varepsilon^{-\frac52}$), $\varepsilon>0$ being a prescribed accuracy, which is suboptimal compared to the canonical multilevel Monte Carlo performance. This suboptimality stems from the discontinuity ofthe Heaviside function involved in the biased stochastic gradient that is recursively evaluated to derive the value-at-risk. To mitigate this issue, this paper proposes and analyzes a multilevel stochastic approximation algorithm that adaptively selects the number of inner samples at each level, and proves that its best complexity is in O($\varepsilon^{-2}|\ln{\varepsilon}|^\frac52$). Our theoretical analysis is exemplified through numerical experiments.

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

1 major / 1 minor

Summary. The manuscript extends the multilevel stochastic approximation (SA) scheme of Crépey, Frikha, and Louzi (2025) for Monte Carlo estimation of Value-at-Risk (VaR). It introduces an adaptive rule that selects the number of inner samples at each level to offset bias induced by the discontinuous Heaviside function in the stochastic gradient. Under stated regularity conditions on the loss random variable, the authors prove that the resulting algorithm attains a complexity of O(ε^{-2} |ln ε|^{5/2}), improving on the prior O(ε^{-5/2}) rate. Numerical experiments are presented to support the analysis.

Significance. If the complexity bound holds under the stated assumptions, the result would represent a concrete advance in the computational cost of multilevel SA methods for VaR, a core quantity in quantitative risk management. The adaptive inner-sample mechanism directly targets the source of suboptimality identified in the base scheme. The inclusion of numerical experiments provides some empirical grounding, though their scope is not detailed in the abstract.

major comments (1)
  1. [Abstract] Abstract (paragraph on regularity conditions): The improved O(ε^{-2} |ln ε|^{5/2}) complexity is derived under regularity conditions on the loss random variable that enable the adaptive inner-sample rule to keep the bias from the Heaviside discontinuity at O(ε) uniformly across levels. The manuscript does not appear to verify that these conditions (e.g., density smoothness or tail behavior) are satisfied for the loss distributions typically arising in financial applications, nor does it quantify the degradation if the density is merely continuous rather than Lipschitz. This assumption is load-bearing for the claimed rate improvement.
minor comments (1)
  1. [Abstract] The abstract refers to 'numerical experiments' but provides no information on the test cases, error controls, or data-selection rules used; this should be expanded in the main text for reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive comment on the regularity conditions. We address it below and will incorporate clarifications in the revision.

read point-by-point responses

  1. Referee: The improved O(ε^{-2} |ln ε|^{5/2}) complexity is derived under regularity conditions on the loss random variable that enable the adaptive inner-sample rule to keep the bias from the Heaviside discontinuity at O(ε) uniformly across levels. The manuscript does not appear to verify that these conditions (e.g., density smoothness or tail behavior) are satisfied for the loss distributions typically arising in financial applications, nor does it quantify the degradation if the density is merely continuous rather than Lipschitz. This assumption is load-bearing for the claimed rate improvement.

    Authors: We agree that the Lipschitz continuity of the density (Assumption 2.1) is essential for the uniform O(ε) bias control that yields the improved rate. The manuscript states these conditions explicitly but does not include verification for common financial losses or sensitivity analysis under weaker (continuous but non-Lipschitz) densities. In the revision we will add Remark 2.2 with examples (e.g., Black-Scholes loss, which satisfies the Lipschitz condition via the Gaussian density) and a brief discussion noting that under mere continuity the bias control weakens, likely reverting the complexity to the non-adaptive O(ε^{-5/2}) rate; we will not claim a precise intermediate rate without further analysis. revision: yes

Circularity Check

0 steps flagged

Minor self-citation to prior base scheme; adaptive extension and complexity bound are independent

full rationale

The paper cites Crépey, Frikha, and Louzi (2025) for the non-adaptive multilevel SA scheme and its O(ε^{-5/2}) complexity, noting the source of suboptimality in the Heaviside discontinuity. The present work then introduces a new adaptive inner-sample selection rule and derives a distinct O(ε^{-2}|ln ε|^{5/2}) bound under stated regularity assumptions on the loss variable. This extension does not reduce by construction to the prior result; the adaptive bias control and complexity proof constitute independent content. No self-definitional equations, fitted inputs renamed as predictions, or load-bearing uniqueness theorems from overlapping authors are present. The self-citation is therefore minor and non-circular per the guidelines.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no explicit free parameters, axioms, or invented entities; the analysis presumably inherits standard assumptions on the loss distribution and Monte Carlo simulation from the cited 2025 paper.

pith-pipeline@v0.9.0 · 5689 in / 1096 out tokens · 43938 ms · 2026-05-23T21:43:25.229070+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. Mission-Aligned Learning-Informed Control of Autonomous Systems: Formulation and Foundations

    math.OC 2025-07 unverdicted novelty 4.0

    The paper formulates a two-level optimization scheme integrating control, classical planning, and reinforcement learning to improve safety and interpretability in autonomous systems.

Reference graph

Works this paper leans on

33 extracted references · 33 canonical work pages · cited by 1 Pith paper · 1 internal anchor

  1. [1]

    On the coherence of expected shortfall

    C. Acerbi and D. Tasche. “On the coherence of expected shortfall”. In:Journal of Banking & Finance26.7 (2002), pp. 1487–1503

    work page 2002

  2. [2]

    Computing VaR and CVaR using stochastic approx- imation and adaptive unconstrained importance sampling

    O. Bardou, N. Frikha, and G. Pagès. “Computing VaR and CVaR using stochastic approx- imation and adaptive unconstrained importance sampling”. In:Monte Carlo Methods and Applications 15.3 (2009), pp. 173–210

    work page 2009

  3. [3]

    CVaR hedging using quantization-based stochastic approximation algorithm

    O. Bardou, N. Frikha, and G. Pagès. “CVaR hedging using quantization-based stochastic approximation algorithm”. In:Mathematical Finance26.1 (2016), pp. 184–229

    work page 2016

  4. [4]

    Recursive computation of value-at-risk and condi- tional value-at-risk using MC and QMC

    O. Bardou, N. Frikha, and G. Pagès. “Recursive computation of value-at-risk and condi- tional value-at-risk using MC and QMC”. In:Monte Carlo and Quasi-Monte Carlo Methods

    work page

  5. [5]

    Springer, 2009, pp. 193–208

    work page 2009

  6. [6]

    Stochastic ap- proximation schemes for economic capital and risk margin computations

    D. Barrera, S. Crépey, B. Diallo, G. Fort, E. Gobet, and U. Stazhynski. “Stochastic ap- proximation schemes for economic capital and risk margin computations”. In:ESAIM: Proceedings and Surveys65 (2019), pp. 182–218

    work page 2019

  7. [7]

    Basel Committee on Banking Supervision.Consultative document: Fundamental Review of the Trading Book: A revised market risk framework. 2013. url: bis.org/publ/bcbs265. pdf

    work page 2013

  8. [8]

    Adaptive importance sampling for multilevel Monte Carlo Euler method

    M. Ben Alaya, K. Hajji, and A. Kebaier. “Adaptive importance sampling for multilevel Monte Carlo Euler method”. In:Stochastics 95.2 (2023), pp. 303–327

    work page 2023

  9. [9]

    An old-new concept of convex risk measures: the optimized certainty equivalent

    A. Ben-Tal and M. Teboulle. “An old-new concept of convex risk measures: the optimized certainty equivalent”. In:Mathematical Finance17.3 (2007), pp. 449–476

    work page 2007

  10. [10]

    Stochastic approximation algorithms for superquantiles estimation

    B. Bercu, M. Costa, and S. Gadat. “Stochastic approximation algorithms for superquantiles estimation”. In:Electronic Journal of Probability26 (2021), pp. 1–29

    work page 2021

  11. [11]

    Efficient risk estimation via nested sequential simulation

    M. Broadie, Y. Du, and C. C. Moallemi. “Efficient risk estimation via nested sequential simulation”. In:Management Science57.6 (2011), pp. 1172–1194

    work page 2011

  12. [12]

    MultilevelsimulationoffunctionalsofBernoulli random variables with application to basket credit derivatives

    K.Bujok,B.M.Hambly,andC.Reisinger.“MultilevelsimulationoffunctionalsofBernoulli random variables with application to basket credit derivatives”. In:Methodology and Com- puting in Applied Probability17.3 (2015), pp. 579–604

    work page 2015

  13. [13]

    Non asymptotic controls on a recursive superquantile approxi- mation

    M. Costa and S. Gadat. “Non asymptotic controls on a recursive superquantile approxi- mation”. In:Electronic Journal of Statistics15.2 (2021), pp. 4718–4769

    work page 2021

  14. [14]

    A Multilevel Stochastic Approximation Algorithm for Value-at-Risk and Expected Shortfall Estimation

    S. Crépey, N. Frikha, and A. Louzi. A multilevel stochastic approximation algorithm for value-at-risk and expected shortfall estimation. 2023. arXiv:2304.01207

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  15. [15]

    Crépey, N

    S. Crépey, N. Frikha, A. Louzi, and G. Pagès. Asymptotic error analysis of multilevel stochastic approximations for the value-at-risk and expected shortfall. 2023. arXiv: 2311. 15333

    work page 2023

  16. [16]

    Multilevel Monte Carlo algorithms for Lévy-driven SDEs with Gaussian cor- rection

    S. Dereich. “Multilevel Monte Carlo algorithms for Lévy-driven SDEs with Gaussian cor- rection”. In:The Annals of Applied Probability21.1 (2011), pp. 283–311

    work page 2011

  17. [17]

    General multilevel adaptations for stochastic approx- imation algorithms of Robbins–Monro and Polyak–Ruppert type

    S. Dereich and T. Müller-Gronbach. “General multilevel adaptations for stochastic approx- imation algorithms of Robbins–Monro and Polyak–Ruppert type”. In:Numerische Mathe- matik 142.2 (2019), pp. 279–328

    work page 2019

  18. [18]

    A multilevel Monte Carlo method for comput- ing failure probabilities

    D. Elfverson, F. Hellman, and A. Mälqvist. “A multilevel Monte Carlo method for comput- ing failure probabilities”. In:SIAM/ASA Journal on Uncertainty Quantification4.1 (2016), pp. 312–330

    work page 2016

  19. [19]

    Transport-entropyinequalitiesanddeviationestimatesforstochas- tic approximation schemes

    M.FathiandN.Frikha.“Transport-entropyinequalitiesanddeviationestimatesforstochas- tic approximation schemes”. In:Electronic Journal of Probability18 (2013), pp. 1–36. 42

    work page 2013

  20. [20]

    Convex risk measures

    H. Föllmer and A. Schied. “Convex risk measures”. In:Encyclopedia of Quantitative Fi- nance. John Wiley & Sons, Ltd, 2010

    work page 2010

  21. [21]

    Multi-level stochastic approximation algorithms

    N. Frikha. “Multi-level stochastic approximation algorithms”. In:The Annals of Applied Probability 26.2 (2016), pp. 933–985

    work page 2016

  22. [22]

    Concentration bounds for stochastic approximations

    N. Frikha and S. Menozzi. “Concentration bounds for stochastic approximations”. In:Elec- tronic Communications in Probability17 (2012), pp. 1–15

    work page 2012

  23. [23]

    M. B. Giles. MLMC techniques for discontinuous functions. 2023. arXiv:2301.02882

    work page arXiv 2023

  24. [24]

    Multilevel Monte Carlo path simulation

    M. B. Giles. “Multilevel Monte Carlo path simulation”. In:Operations Research56.3 (2008), pp. 607–617

    work page 2008

  25. [25]

    Multilevel nested simulation for efficient risk estimation

    M. B. Giles and A.-L. Haji-Ali. “Multilevel nested simulation for efficient risk estimation”. In: SIAM/ASA Journal on Uncertainty Quantification7.2 (2019), pp. 497–525

    work page 2019

  26. [26]

    M. B. Giles, A.-L. Haji-Ali, and J. Spence.Efficient risk estimation for the credit valuation adjustment. 2023. arXiv:2301.05886

    work page arXiv 2023

  27. [27]

    Weak error for nested multilevel Monte Carlo

    D. Giorgi, V. Lemaire, and G. Pagès. “Weak error for nested multilevel Monte Carlo”. In: Methodology and Computing in Applied Probability22 (3 2020), pp. 1325–1348

    work page 2020

  28. [28]

    Nested simulation in portfolio risk measurement

    M. B. Gordy and S. Juneja. “Nested simulation in portfolio risk measurement”. In:Man- agement Science56.10 (2010), pp. 1833–1848

    work page 2010

  29. [29]

    Adaptive multilevel Monte Carlo for probabilities

    A.-L. Haji-Ali, J. Spence, and A. L. Teckentrup. “Adaptive multilevel Monte Carlo for probabilities”. In:SIAM Journal on Numerical Analysis60.4 (2022), pp. 2125–2149

    work page 2022

  30. [30]

    Monte Carlo complexity of global solution of integral equations

    S. Heinrich. “Monte Carlo complexity of global solution of integral equations”. In:Journal of Complexity 14.2 (1998), pp. 151–175

    work page 1998

  31. [31]

    MultilevelMonteCarlomethods

    S.Heinrich.“MultilevelMonteCarlomethods”.In: Large-Scale Scientific Computing.Springer, 2001, pp. 58–67

    work page 2001

  32. [32]

    Implementation and analysis of an adaptive multilevel Monte Carlo algorithm

    H. Hoel, E. von Schwerin, A. Szepessy, and R. Tempone. “Implementation and analysis of an adaptive multilevel Monte Carlo algorithm”. In:Monte Carlo Methods and Applications 20.1 (2014), pp. 1–41

    work page 2014

  33. [33]

    Optimization of conditional value-at-risk

    R. T. Rockafellar and S. Uryasev. “Optimization of conditional value-at-risk”. In:Journal of Risk 2.3 (2000), pp. 21–41. 43

    work page 2000