pith. sign in

arxiv: 2605.19978 · v1 · pith:TZQXDLDFnew · submitted 2026-05-19 · 🧮 math.OC · math.AP· math.PR

Analytical Approach to Continuous-Time Causal Optimal Transport

Julio Backhoff , Erhan Bayraktar , Ibrahim Ekren , Antonios Zitridis This is my paper

Pith reviewed 2026-05-20 03:40 UTC · model grok-4.3

classification 🧮 math.OC math.APmath.PR
keywords causal optimal transportcontinuous timemaster equationstochastic controlKushner-Stratonovich filteringMarkov processdiffusion process
0
0 comments X

The pith

The value of continuous-time causal optimal transport equals the solution of a fully nonlinear parabolic master equation obtained by conditioning the source on the target.

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

The paper examines causal optimal transport in continuous time between a finite-state Markov source and a diffusion target when the cost is Markovian. It shows that the problem can be reformulated by replacing the source process with the conditional distribution of its state given observations of the target. This reformulation leads to a characterization of the transport value via a fully nonlinear parabolic master equation defined on an enlarged state space. The same value is shown to coincide with the values of two stochastic control problems, one controlling the filtering equation and another with state constraints on the simplex. These equivalences provide implementable numerical schemes that bound the value from above and below.

Core claim

By replacing the source with its conditional law given the observation of the target, the value of this transport problem is characterized through a fully nonlinear parabolic master equation on an enlarged state space, and this value coincides with those of two equivalent stochastic control problems on the simplex: a control of the Kushner-Stratonovich filtering equation with a zero-mean condition and a state-constrained stochastic optimal control problem.

What carries the argument

The conditional law of the Markov source given observations of the diffusion target, which serves as the new state variable to formulate the master equation and the control problems.

If this is right

  • The transport value satisfies the master equation and can be solved numerically from it.
  • Equivalent control problems on the simplex allow approximation of the value from both above and below.
  • The formulation connects causal transport directly to filtering equations and constrained optimization.
  • Both control formulations give rise to implementable numerical schemes.

Where Pith is reading between the lines

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

  • This method may allow computation of causal transport costs in applications like stochastic filtering or time-series analysis where direct optimization is hard.
  • Extensions could involve relaxing the Markovian assumptions or applying similar conditioning to other transport problems.
  • The numerical schemes could be validated on low-dimensional examples to confirm the equivalences hold in practice.

Load-bearing premise

The replacement of the source by its conditional law given target observations preserves the value of the causal transport problem and leads to a well-posed master equation, which requires sufficient regularity on the Markov source, diffusion target, and Markovian cost.

What would settle it

For a simple case with a two-state Markov chain as source and standard Brownian motion as target, compute the causal transport cost directly and verify whether it equals the numerical solution of the master equation or the control problems.

read the original abstract

We study causal optimal transport in continuous time, with Markovian cost, between a finite-state Markov source and a diffusion target. By replacing the source with its conditional law given the observation of the target, we characterize the value of this transport problem through a fully nonlinear parabolic master equation on an enlarged state space. We further show that this value coincides with those of two equivalent stochastic control problems on the simplex: a control of the Kushner--Stratonovich filtering equation with a zero-mean condition, and a state-constrained stochastic optimal control problem. Both formulations give rise to implementable numerical schemes that approximate the value from above and below.

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

0 major / 3 minor

Summary. The paper develops an analytical framework for continuous-time causal optimal transport between a finite-state Markov source and a diffusion target under a Markovian cost. By replacing the source with its conditional law given target observations, the value is characterized via a fully nonlinear parabolic master equation on an enlarged state space (probability simplex plus target state). Equivalences are established to two stochastic control problems: control of the Kushner-Stratonovich equation with a zero-mean constraint, and a state-constrained stochastic optimal control problem. These yield implementable numerical schemes providing upper and lower approximations.

Significance. If the derivations hold, the work is significant for rigorously connecting causal optimal transport to nonlinear filtering and stochastic control on the simplex. Strengths include the use of standard filtering theory and dynamic programming without hidden circularity, explicit regularity assumptions on transition rates, diffusion coefficients, and costs, and the provision of concrete numerical schemes. This offers both theoretical insight and practical tools for problems in stochastic optimization.

minor comments (3)
  1. §2.2: The definition of the enlarged state space (conditional law μ on the simplex together with target state x) is introduced after the master equation is stated; moving the notation and state-space description earlier would improve readability of the subsequent derivations.
  2. §4.1, Eq. (18): The zero-mean condition on the control is stated but its precise relation to the Kushner-Stratonovich dynamics could be cross-referenced to the filtering literature (e.g., a brief remark on the innovation process) for readers less familiar with the filtering-control equivalence.
  3. The numerical schemes in §5 are described at a high level; adding a short pseudocode outline or explicit discretization step for the simplex would make the 'implementable' claim more concrete without lengthening the paper.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment of our manuscript and for recommending minor revision. We appreciate the recognition of the paper's contributions in linking causal optimal transport to nonlinear filtering and stochastic control.

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper's central derivation replaces the Markov source with its conditional law given target observations to obtain a master equation on the enlarged state space (conditional law on the simplex plus target state), then equates the value to two equivalent stochastic control problems via the Kushner-Stratonovich equation and state-constrained control. These steps rely on standard filtering theory and dynamic programming under explicitly stated regularity assumptions on transition rates, diffusion coefficients, and costs; no equation reduces to its inputs by construction, no fitted parameters are relabeled as predictions, and no load-bearing self-citations or imported uniqueness theorems appear. The argument is self-contained against external benchmarks from stochastic control and filtering.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard domain assumptions from stochastic processes and optimal transport; no free parameters, new invented entities, or ad-hoc axioms are introduced in the abstract.

axioms (2)
  • domain assumption The source is a finite-state Markov process and the target is a diffusion process with Markovian cost.
    Stated directly in the problem setup of the abstract.
  • domain assumption The conditional law of the source given target observations can be used to enlarge the state space while preserving the transport value.
    Central modeling step invoked to derive the master equation.

pith-pipeline@v0.9.0 · 5638 in / 1501 out tokens · 46292 ms · 2026-05-20T03:40:07.686873+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages · 2 internal anchors

  1. [1]

    The geometry of the adapted bures–wasserstein space.arXiv preprint arXiv:2602.00623,

    [ABG+26] Beatrice Acciaio, Daniel Bartl, Anne Grass, Songyan Hou, and Gudmund Pammer. The geometry of the adapted bures–wasserstein space.arXiv preprint arXiv:2602.00623,

    work page arXiv

  2. [2]

    Causal transport in discrete time and applications

    [BBLZ16] Julio Backhoff, Mathias Beiglb¨ ock, Yiqing Lin, and Anastasiia Zalashko. Causal transport in discrete time and applications.arXiv preprint arXiv:1606.04062,

    work page internal anchor Pith review Pith/arXiv arXiv

  3. [3]

    The Wasserstein space of stochastic pro- cesses.arXiv preprint arXiv:2104.14245, pages 1–41,

    [BBP21] Daniel Bartl, Mathias Beiglb¨ ock, and Gudmund Pammer. The Wasserstein space of stochastic pro- cesses.arXiv preprint arXiv:2104.14245, pages 1–41,

    work page arXiv

  4. [4]

    The Wasserstein space of stochastic processes in continuous time.arXiv preprint arXiv:2501.14135,

    [BBP+25] Daniel Bartl, Mathias Beiglb¨ ock, Gudmund Pammer, Stefan Schrott, and Xin Zhang. The Wasserstein space of stochastic processes in continuous time.arXiv preprint arXiv:2501.14135,

    work page arXiv

  5. [5]

    Optimal transport and risk aversion in Kyle’s model of informed trading.arXiv e-prints, pages arXiv–2006,

    [BCEL20] Kerry Back, Francois Cocquemas, Ibrahim Ekren, and Abraham Lioui. Optimal transport and risk aversion in Kyle’s model of informed trading.arXiv e-prints, pages arXiv–2006,

    work page 2006

  6. [6]

    Kyle-Back models with risk aversion and non-Gaussian beliefs.arXiv preprint arXiv:2008.06377,

    [BE20] Shreya Bose and Ibrahim Ekren. Kyle-Back models with risk aversion and non-Gaussian beliefs.arXiv preprint arXiv:2008.06377,

    work page arXiv 2008

  7. [7]

    Multidimensional Kyle-Back model with a risk averse informed trader

    ANALYTICAL APPROACH TO CONTINUOUS-TIME CAUSAL OPTIMAL TRANSPORT 51 [BE21] Shreya Bose and Ibrahim Ekren. Multidimensional Kyle-Back model with a risk averse informed trader. arXiv preprint arXiv:2111.01957,

    work page arXiv

  8. [8]

    Goal-based portfolio selection with mental accounting

    [BH25b] Erhan Bayraktar and Bingyan Han. Goal-based portfolio selection with mental accounting.arXiv preprint arXiv:2506.06654,

    work page internal anchor Pith review Pith/arXiv arXiv

  9. [9]

    Bounding adapted Wasserstein metrics.arXiv preprint arXiv:2407.21492,

    [BLPW24] Jose Blanchet, Martin Larsson, Jonghwa Park, and Johannes Wiesel. Bounding adapted Wasserstein metrics.arXiv preprint arXiv:2407.21492,

    work page arXiv

  10. [10]

    Sensitivity of multiperiod optimization problems in adapted Wasser- stein distance.arXiv preprint arXiv:2208.05656,

    [BW22] Daniel Bartl and Johannes Wiesel. Sensitivity of multiperiod optimization problems in adapted Wasser- stein distance.arXiv preprint arXiv:2208.05656,

    work page arXiv

  11. [11]

    BACKHOFF, E

    52 J. BACKHOFF, E. BAYRAKTAR, I. EKREN, AND A. ZITRIDIS [CENV22] Reda Chhaibi, Ibrahim Ekren, Eunjung Noh, and Lu Vy. A unified approach to informed trading via Monge-Kantorovich duality.https://arxiv.org/pdf/2210.17384.pdf,

    work page arXiv

  12. [12]

    Causal transport on path space.arXiv preprint arXiv:2412.02948,

    [CL24] Rama Cont and Fang Rui Lim. Causal transport on path space.arXiv preprint arXiv:2412.02948,

    work page arXiv

  13. [13]

    Bicausal optimal transport for sdes with irregular coefficients

    [HR24] Michaela Hitz and Benjamin A Robinson. Bicausal optimal transport for sdes with irregular coefficients. arXiv preprint arXiv:2403.09941,

    work page arXiv

  14. [14]

    Duality of causal distributionally robust optimization.arXiv preprint arXiv:2401.16556,

    [Jia24] Yifan Jiang. Duality of causal distributionally robust optimization.arXiv preprint arXiv:2401.16556,

    work page arXiv

  15. [15]

    A transfer principle for computing the adapted wasserstein distance between stochastic processes.arXiv preprint arXiv:2505.21337,

    [JL25] Yifan Jiang and Fang Rui Lim. A transfer principle for computing the adapted wasserstein distance between stochastic processes.arXiv preprint arXiv:2505.21337,

    work page arXiv

  16. [16]

    Sensitivity of causal distributionally robust optimization.arXiv preprint arXiv:2408.17109,

    [JO24] Yifan Jiang and Jan Obloj. Sensitivity of causal distributionally robust optimization.arXiv preprint arXiv:2408.17109,

    work page arXiv

  17. [17]

    Viscosity solutions of fully nonlinear second order equations and optimal stochastic control in infinite dimensions

    [Lio89b] Pierre-Louis Lions. Viscosity solutions of fully nonlinear second order equations and optimal stochastic control in infinite dimensions. part ii: Optimal control of Zakai’s equation. InStochastic Partial Differ- ential Equations and Applications II: Proceedings of a Conference held in Trento, Italy February 1–6, 1988, pages 147–170. Springer,

    work page 1988

  18. [18]

    The fast rate of convergence of the smooth adapted Wasserstein distance.arXiv preprint arXiv:2503.10827,

    [LPW25] Martin Larsson, Jonghwa Park, and Johannes Wiesel. The fast rate of convergence of the smooth adapted Wasserstein distance.arXiv preprint arXiv:2503.10827,

    work page arXiv

  19. [19]

    Robust optimization in causal models and g-causal normalizing flows.arXiv preprint arXiv:2510.15458,

    [VC25] Gabriele Visentin and Patrick Cheridito. Robust optimization in causal models and g-causal normalizing flows.arXiv preprint arXiv:2510.15458,

    work page arXiv