pith. sign in

arxiv: 2507.18443 · v2 · submitted 2025-07-24 · 🧮 math.NA · cs.NA

On MAP estimates and source conditions for drift identification in SDEs

Daniel Tenbrinck , Nikolas Uesseler , Philipp Wacker , Benedikt Wirth This is my paper

Pith reviewed 2026-05-19 02:43 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords drift identificationstochastic differential equationsMAP estimatetangential cone conditioninverse problemsconvergence ratessource conditionsnumerical simulations
0
0 comments X

The pith

A MAP estimate for identifying the drift in stochastic differential equations from discrete observations satisfies a tangential cone condition on the forward operator.

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

The paper establishes a maximum a posteriori estimate for recovering the drift term of an SDE given multiple observations of its path at fixed times. It proves that the forward mapping from drift to observations is differentiable and obeys a tangential cone condition, which is a key property for analyzing inverse problems. This sets the stage for applying existing convergence theory to show that the estimate improves as the number of observation sets n grows to infinity. A sympathetic reader would care because accurate drift recovery enables better modeling and prediction in systems governed by stochastic dynamics, such as in physics or finance. Simulations provide initial support that the rates hold in practice.

Core claim

We consider the inverse problem of identifying the drift in an SDE from n observations of its solution at M+1 distinct time points. We derive a corresponding MAP estimate, we prove differentiability properties as well as a so-called tangential cone condition for the forward operator, and we review the existing theory for related problems, which under a slightly stronger tangential cone condition would additionally yield convergence rates for the MAP estimate as n→∞. Numerical simulations in 1D indicate that such convergence rates indeed hold true.

What carries the argument

The tangential cone condition for the forward operator, which relates the distance between the operator applied to two points to the distance in parameter space and supports stability analysis for the inverse problem.

If this is right

  • The MAP estimate provides a practical method to identify the drift coefficient from discrete trajectory data.
  • Differentiability of the forward operator allows for gradient-based optimization in computing the estimate.
  • The tangential cone condition holds for this setup, facilitating the use of regularization theory.
  • Under a slightly stronger version of the condition, convergence rates for the MAP estimate can be obtained as the number of observations increases.
  • 1D numerical simulations suggest that the convergence rates are achieved in practice.

Where Pith is reading between the lines

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

  • This framework could be applied to multi-dimensional SDEs where analytical proofs are harder but simulations can guide.
  • Connecting to source conditions might allow for improved regularization strategies in similar inverse problems for stochastic processes.
  • Extensions to continuous observations or different noise models could follow from the same operator properties.

Load-bearing premise

The forward operator from drift functions to observation data must satisfy a tangential cone condition, which the paper proves but requires a slightly stronger version for full convergence guarantees.

What would settle it

Perform simulations with increasing n in one dimension and check if the error between the estimated drift and true drift decreases at the predicted rate; failure to observe the rate would question the applicability of the stronger condition.

Figures

Figures reproduced from arXiv: 2507.18443 by Benedikt Wirth, Daniel Tenbrinck, Nikolas Uesseler, Philipp Wacker.

Figure 1
Figure 1. Figure 1: Left: Epifluorescence microscopy data of a zebrafish embryo (PGCs in red). Right: PGC distribution [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The sketch illustrates that for smaller δ the minimal value of S(·, yδ ) approaches S(y † , yδ ). Thus for small δ, y † almost minimizes S(·, yδ ). Depending on the specific properties of S, this implies information on how close y δ is to y † . For rates of convergence we need to generalize the previous variational source conditions. Assumption 3.2 (Conditions for convergence rates VII, [19]). 1. Let ϕ be … view at source ↗
Figure 3
Figure 3. Figure 3: Left: Simulated trajectories (top) and potential inference (bottom) for [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Error between ground truth and inferred potential in [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
read the original abstract

We consider the inverse problem of identifying the drift in an SDE from $n$ observations of its solution at $M+1$ distinct time points. We derive a corresponding MAP estimate, we prove differentiability properties as well as a so-called tangential cone condition for the forward operator, and we review the existing theory for related problems, which under a slightly stronger tangential cone condition would additionally yield convergence rates for the MAP estimate as $n\to\infty$. Numerical simulations in 1D indicate that such convergence rates indeed hold true.

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 / 2 minor

Summary. The paper considers the inverse problem of recovering the drift function in a stochastic differential equation from n discrete observations of solution paths at M+1 time points. It derives the corresponding MAP estimator, establishes differentiability of the forward operator, and proves that the operator satisfies a tangential cone condition. The authors review convergence theory from related inverse problems and note that a modestly stronger tangential cone condition would imply rates for the MAP estimator as n→∞; 1D numerical experiments are presented as supporting evidence that the rates appear to hold in practice.

Significance. If the stronger tangential cone condition can be established or if the numerical indication generalizes, the work supplies a technically grounded MAP framework for drift identification in SDEs together with the first verification of the tangential cone condition in this setting. The explicit proof of the (standard) tangential cone condition is a concrete technical contribution that could serve as a stepping stone for rate results once the stronger variant is settled.

major comments (2)
  1. [§4] §4 (review of convergence theory): the statement that existing theory would deliver convergence rates under a 'slightly stronger' tangential cone condition is left conditional; the manuscript neither proves nor verifies this stronger condition for the SDE forward operator and instead cites 1D simulations as indicative evidence. Because the rate claim is presented as a principal motivation, the gap between the proven tangential cone condition and the stronger variant required by the cited theory is load-bearing for the convergence-rate assertion.
  2. [§5] §5 (numerical experiments): the simulations are restricted to one dimension and do not report error bars, multiple random seeds, or quantitative comparison against the predicted rates; this limits the strength of the numerical support for the claim that the stronger tangential cone condition 'indeed holds true' in the setting of the paper.
minor comments (2)
  1. [§3] Clarify the precise functional setting (e.g., the Banach space for the drift and the precise observation operator) when the tangential cone condition is stated; the current notation leaves the precise norms implicit.
  2. [§5] Add a short discussion of how the MAP estimator is computed in practice (optimization algorithm, discretization of the SDE) so that the numerical results can be reproduced.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their thoughtful review and valuable suggestions. We address each major comment below and outline the revisions we intend to implement.

read point-by-point responses

  1. Referee: [§4] §4 (review of convergence theory): the statement that existing theory would deliver convergence rates under a 'slightly stronger' tangential cone condition is left conditional; the manuscript neither proves nor verifies this stronger condition for the SDE forward operator and instead cites 1D simulations as indicative evidence. Because the rate claim is presented as a principal motivation, the gap between the proven tangential cone condition and the stronger variant required by the cited theory is load-bearing for the convergence-rate assertion.

    Authors: We acknowledge that our discussion of convergence rates is conditional on a stronger tangential cone condition, which we have not established analytically. The manuscript reviews the relevant theory from the literature and uses 1D numerical experiments to provide supporting evidence that the rates appear to hold. In revision, we will clarify in the abstract, introduction, and conclusion that the rates are not proven but suggested by the numerics, and we will add a brief discussion highlighting the verification of the stronger condition as an open question for future work. This adjustment ensures the claims are appropriately qualified while preserving the technical contribution of proving the standard tangential cone condition. revision: yes

  2. Referee: [§5] §5 (numerical experiments): the simulations are restricted to one dimension and do not report error bars, multiple random seeds, or quantitative comparison against the predicted rates; this limits the strength of the numerical support for the claim that the stronger tangential cone condition 'indeed holds true' in the setting of the paper.

    Authors: The numerical section is intended as an initial illustration in one dimension to demonstrate the practical behavior of the estimator as the number of observations increases. We agree that additional statistical rigor would strengthen the presentation. In the revised manuscript, we will include results averaged over multiple independent random seeds, report error bars or standard deviations, and provide a quantitative assessment comparing the observed convergence rates to those predicted by the theory under the stronger condition. We will also consider extending to a simple two-dimensional example if space permits. revision: yes

standing simulated objections not resolved
  • Proving the stronger tangential cone condition for the SDE drift identification problem, which remains an open analytical challenge beyond the scope of the current work.

Circularity Check

0 steps flagged

No circularity: MAP derivation and tangential cone proof are independent of inputs

full rationale

The paper derives the MAP estimate directly, proves differentiability properties and the tangential cone condition for the forward operator from first principles, and reviews external theory for rates under a stronger condition without claiming to establish that stronger condition analytically. Numerical simulations are presented only as supporting indication rather than as the definition of any result. No quoted step reduces a claimed prediction or theorem to a fitted parameter, self-citation chain, or input by construction; the central claims remain self-contained against the stated assumptions.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities are identifiable from the given text.

pith-pipeline@v0.9.0 · 5617 in / 1177 out tokens · 81273 ms · 2026-05-19T02:43:06.470812+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

19 extracted references · 19 canonical work pages

  1. [1]

    An extension of the variational inequality approach for nonlinear ill-posed problems

    Radu Bo¸ t and Bernd Hofmann. An extension of the variational inequality approach for nonlinear ill-posed problems. 09 2009. 14 Number n of particles L2-norm error O(n−1/2) O(n−1) Figure 4: Error between ground truth and inferred potential in L2 norm, as a function of number of particles tracked, visualised as a box plot over 25 independent runs each. Asy...

    work page 2009

  2. [2]

    J. M. Borwein and A. S. Lewis. Convergence of best entropy estimates. SIAM Journal on Optimization , 1(2):191–205, 1991

    work page 1991

  3. [3]

    Convergence rates of convex variational regularization

    Martin Burger and Stanley Osher. Convergence rates of convex variational regularization. Inverse Problems, 20, 07 2004

    work page 2004

  4. [4]

    On parameter identification in stochastic differential equations by penalized maximum likelihood

    Fabian Dunker and Thorsten Hohage. On parameter identification in stochastic differential equations by penalized maximum likelihood. Inverse Problems, 30, 08 2014

    work page 2014

  5. [5]

    Convergence rates for tikhonov regularisation of non-linear ill-posed problems

    H Engl, K Kunisch, and A Neubauer. Convergence rates for tikhonov regularisation of non-linear ill-posed problems. Inverse Problems, 5:523, 01 1999

    work page 1999

  6. [6]

    Generalized Tikhonov regularization and modern convergence rate theory in Banach spaces

    Jens Flemming. Generalized Tikhonov regularization and modern convergence rate theory in Banach spaces. Berichte aus der Mathematik. Shaker Verlag, Aachen, 2012

    work page 2012

  7. [7]

    Convergence rates in constrained Tikhonov regularization: equivalence of projected source conditions and variational inequalities

    Jens Flemming and Bernd Hofmann. Convergence rates in constrained Tikhonov regularization: equivalence of projected source conditions and variational inequalities. Inverse Problems, 27(8):085001, 11, 2011

    work page 2011

  8. [8]

    Friedman

    A. Friedman. Partial Differential Equations of Parabolic Type. Dover Books on Mathematics. Dover Publi- cations, 2008

    work page 2008

  9. [9]

    Gross-Thebing, L

    S. Gross-Thebing, L. Truszkowski, D. Tenbrinck, H. Sanchez-Iranzo, C. Camelo, K.J. Westerich, A. Singh, P. Maier, J. Prengel, P. Lange, J. Huewel, F. Gaede, R. Sasse, B.E. V os, T. Betz, M. Matis, R. Prevedel, S. Luschnig, A. Diz-Munoz, M. Burger, and E. Raz. Using migrating cells as probes to illuminate features in live embryonic tissues. Science Advance...

    work page 2020

  10. [10]

    A convergence rates result for tikhonov regularization in banach spaces with non-smooth operators

    Bernd Hofmann, Barbara Kaltenbacher, Christiane Pöschl, and O Scherzer. A convergence rates result for tikhonov regularization in banach spaces with non-smooth operators. Inverse Problems, 23:987, 04 2007

    work page 2007

  11. [11]

    On the interplay of source conditions and variational inequalities for nonlinear ill-posed problems

    Bernd Hofmann and Masahiro Yamamoto. On the interplay of source conditions and variational inequalities for nonlinear ill-posed problems. Applicable Analysis, 89:1705–1727, 11 2010

    work page 2010

  12. [12]

    A. M. Il’in, A. S. Kalashnikov, and O. A. Oleinik. Linear equations of the second order of parabolic type. Uspekhi Matematicheskikh Nauk, 17(3):3–146, 1962

    work page 1962

  13. [13]

    Ladyženskaja, Vsevolod A

    Ol’ga A. Ladyženskaja, Vsevolod A. Solonnikov, and Nina N. Ural’ceva. Linear and Quasi-linear Equa- tions of Parabolic Type, volume 23 of Translations of Mathematical Monographs. American Mathematical Society, 1968

    work page 1968

  14. [14]

    How general are general source conditions? Inverse Problems, 24:015009, 01 2008

    Peter Mathé and Bernd Hofmann. How general are general source conditions? Inverse Problems, 24:015009, 01 2008. 15

    work page 2008

  15. [15]

    Tikhonov Regularization with General Residual Term

    Christiane Pöschl. Tikhonov Regularization with General Residual Term. PhD thesis, 10 2008

    work page 2008

  16. [16]

    Error estimates for non-quadratic regularization and the relation to enhancement

    Elena Resmerita and Otmar Scherzer. Error estimates for non-quadratic regularization and the relation to enhancement. Inverse Problems - INVERSE PROBL, 22, 11 2006

    work page 2006

  17. [17]

    Seidman and Curtis R

    Thomas I. Seidman and Curtis R. V ogel. Well-posedness and convergence of some regularisation methods for nonlinear ill posed problems. Inverse Problems, 5(2):227–238, 1989

    work page 1989

  18. [18]

    Inverse Problems with Poisson Data: Tikhonov-Type Regularization and Iteratively Regular- ized Newton Methods

    Frank Werner. Inverse Problems with Poisson Data: Tikhonov-Type Regularization and Iteratively Regular- ized Newton Methods. PhD thesis, University of Göttingen, Göttingen, Germany, 2012. Doctoral dissertation

    work page 2012

  19. [19]

    Convergence rates in expectation for tikhonov-type regularization of inverse problems with poisson data

    Frank Werner and Thorsten Hohage. Convergence rates in expectation for tikhonov-type regularization of inverse problems with poisson data. Inverse Problems, 28, 10 2012. 16

    work page 2012