pith. sign in

arxiv: 2509.09585 · v3 · submitted 2025-09-11 · 💱 q-fin.PM

Causal PDE-Control Models for Dynamic Portfolio Optimization with Latent Drivers

Alejandro Rodriguez Dominguez This is my paper

Pith reviewed 2026-05-18 18:10 UTC · model grok-4.3

classification 💱 q-fin.PM
keywords causal driversPDE controlportfolio optimizationnonlinear filteringprojection-divergence dualityrisk-neutral measuresmartingale representationstructural breaks
0
0 comments X

The pith

Causal PDE-Control Models recover latent drivers via filtering to produce arbitrage-consistent allocations whose stability cost is quantified by projection-divergence duality.

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

Classical portfolio models break down during structural shifts while flexible machine-learning allocations often lose arbitrage consistency and interpretability. The paper introduces Causal PDE-Control Models that combine structural causal drivers, nonlinear filtering, and forward-backward PDE control to generate robust allocation rules under partial information. Driver-conditional risk-neutral measures are built on the observable filtration together with the associated martingale representation, linking pricing, hedging, and portfolio choice in one geometry. A projection-divergence duality then shows that restricting portfolios to the causal driver span selects the feasible allocation closest to the unconstrained optimum under convex divergence, and a causal completeness condition identifies when a finite driver set captures all systematic premia. Markowitz, CAPM/APT, and Black-Litterman appear as limiting cases while reinforcement learning and deep hedging arise as unconstrained approximations.

Core claim

We construct driver-conditional risk-neutral measures on the observable filtration via filtering together with the corresponding martingale representation, linking pricing, hedging, and portfolio choice under a common information set. We further establish a projection-divergence duality showing that restricting portfolios to the causal driver span selects the feasible allocation closest to the unconstrained optimum under a convex divergence, thereby quantifying the stability cost of deviations from the causal manifold, and derive a causal completeness condition identifying when a finite driver span captures systematic premia. Markowitz, CAPM/APT, and Black-Litterman arise as limiting cases,

What carries the argument

The projection-divergence duality, which identifies the causal-driver-span portfolio as the feasible allocation minimizing convex divergence from the unconstrained optimum, together with the causal completeness condition for systematic premia capture.

If this is right

  • Markowitz, CAPM/APT, and Black-Litterman arise as limiting cases of the framework.
  • Reinforcement learning and deep hedging appear as unconstrained approximations within the same pricing-control geometry.
  • Empirical tests on a U.S. equity panel with more than 300 candidate drivers produce higher Sharpe ratios, lower turnover, and more persistent premia than standard benchmarks.
  • The duality quantifies the stability cost of any deviation from the causal manifold.

Where Pith is reading between the lines

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

  • If the duality holds, factor models that stay near the causal span may exhibit lower drawdowns during regime shifts than fully flexible alternatives.
  • The completeness condition suggests a practical test: performance should plateau once the driver span includes all systematic premia, providing a stopping rule for factor selection.
  • The same filtering-plus-PDE geometry could be applied to derivative hedging or credit portfolios where latent drivers are equally relevant.
  • Enforcing the causal restriction may reduce model risk more effectively than post-hoc regularization in machine-learning allocation systems.

Load-bearing premise

Structural causal drivers exist and can be recovered via nonlinear filtering from the observable filtration to construct driver-conditional risk-neutral measures and enable the martingale representation.

What would settle it

In a controlled simulation with known latent drivers, observing that the recovered causal span fails to produce an allocation whose convex divergence from the unconstrained optimum matches the predicted bound would falsify the projection-divergence duality.

Figures

Figures reproduced from arXiv: 2509.09585 by Alejandro Rodriguez Dominguez.

Figure 1
Figure 1. Figure 1: : Star–DAG representation of the Commonality Principle. Drivers [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: : Unconditional mean–variance representations. [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: First conformal map: unconditional → conditional. Pairwise angle proportions (cosines cos θij ) are preserved under conditioning, up to a time–varying scale λ1(t). deep hedging can be interpreted as approximate variants lacking causal pro￾jection or pricing structure. A CPCM begins with latent drivers Ft that represent fundamental sources of variation. Because drivers are only partially observed, a fil￾ter… view at source ↗
Figure 4
Figure 4. Figure 4: Second conformal map: conditional → sensitivity (beta) space. Angle proportions remain invariant (cos α ′ ij (t) = cos αij (t)), with lengths rescaled by λ2(t). θt that determines weights and is projected onto a feasible set (the driver span). Together, filtering, forward evolution, backward control, and pro￾jection generate a portfolio path with instantaneous return pt = θ ⊤ t rt (cf. 3.1.1) and discounte… view at source ↗
Figure 5
Figure 5. Figure 5: : Causal invariance. Intervening on drivers defines driver-specific [PITH_FULL_IMAGE:figures/full_fig_p016_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: : Portfolios are constrained to the driver span [PITH_FULL_IMAGE:figures/full_fig_p017_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: : PDE duality. Forward Fokker–Planck equations describe the [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: : Filter robustness: Particle Filtering (PF) vs. Extended Kalman [PITH_FULL_IMAGE:figures/full_fig_p046_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: : Pareto frontier across soft–PDE weights [PITH_FULL_IMAGE:figures/full_fig_p046_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: : CPCM variants vs. baselines across regimes and [PITH_FULL_IMAGE:figures/full_fig_p047_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: : Sharpe–turnover relation. CPCM variants compress the pos [PITH_FULL_IMAGE:figures/full_fig_p047_11.png] view at source ↗
read the original abstract

Classical portfolio models degrade under structural breaks, whereas flexible machine-learning allocation methods often lack arbitrage consistency and interpretability. We propose Causal PDE-Control Models (CPCMs), a framework that integrates structural causal drivers, nonlinear filtering, and forward-backward PDE control to produce robust and transparent allocation rules under partial information. We construct driver-conditional risk-neutral measures on the observable filtration via filtering together with the corresponding martingale representation, linking pricing, hedging, and portfolio choice under a common information set. We further establish a projection-divergence duality showing that restricting portfolios to the causal driver span selects the feasible allocation closest to the unconstrained optimum under a convex divergence, thereby quantifying the stability cost of deviations from the causal manifold, and derive a causal completeness condition identifying when a finite driver span captures systematic premia. Markowitz, CAPM/APT, and Black-Litterman arise as limiting cases, while reinforcement learning and deep hedging appear as unconstrained approximations within the same pricing-control geometry. Empirically, on a U.S.equity panel with more than 300 candidate drivers, CPCM solvers achieve higher Sharpe ratios, lower turnover, and more persistent premia than econometric and machine-learning benchmarks.

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

3 major / 2 minor

Summary. The manuscript proposes Causal PDE-Control Models (CPCMs) that integrate structural causal drivers recovered via nonlinear filtering, forward-backward PDE control, and martingale representation on the observable filtration to derive dynamic portfolio rules under partial information. It claims a projection-divergence duality under which restriction to the causal driver span yields the minimum-distance allocation to the unconstrained optimum for a convex divergence, together with a causal completeness condition for finite spans capturing systematic premia; classical models (Markowitz, CAPM/APT, Black-Litterman) emerge as limits and empirical results on a U.S. equity panel with >300 candidate drivers report higher Sharpe ratios and lower turnover than benchmarks.

Significance. If the filtering step preserves measure equivalence and the predictable representation property, the framework would supply a unified, arbitrage-consistent geometry linking causal inference, stochastic control, and portfolio optimization, with explicit quantification of the stability cost of deviating from the causal manifold. The empirical outperformance and reduction to classical cases would then constitute a substantive advance in robust allocation under latent drivers.

major comments (3)
  1. [Framework construction (abstract and §3)] Abstract and framework construction: the driver-conditional risk-neutral measures are obtained by nonlinear filtering of latent structural drivers onto the observable filtration, after which a martingale representation is invoked to equate pricing, hedging, and control. No explicit conditions on the observation process, identifiability of the drivers, or preservation of the predictable representation property are stated; without them the subsequent projection-divergence duality may select a projection that is not the true minimum-distance allocation under the original measure.
  2. [Projection-divergence duality (abstract and §4)] Projection-divergence duality: the claim that the causal driver span selects the feasible allocation closest to the unconstrained optimum under a convex divergence presupposes that the filtered measure remains equivalent to the original risk-neutral measure and that the driver span is closed in the relevant L2 space. The manuscript must supply the precise statement, the convexity assumption on the divergence, and the verification that the duality is not circular with the filtering step.
  3. [Causal completeness condition (abstract and §5)] Causal completeness condition: the condition identifying when a finite driver span captures systematic premia is load-bearing for the claim that CPCMs generalize classical models. The empirical selection among >300 candidate drivers introduces a free parameter whose effect on the completeness threshold must be quantified; otherwise the reported persistence of premia may be driven by post-selection bias rather than the causal structure.
minor comments (2)
  1. [Abstract] The abstract states theoretical constructs and empirical outperformance but provides no derivation outline, error analysis, or data-exclusion protocol; a short technical appendix summarizing the key martingale-representation step would improve verifiability.
  2. [Empirical results] Empirical section: report the precise number of drivers retained after any filtering or regularization step, the out-of-sample periods used, and turnover statistics with standard errors to allow direct comparison with the machine-learning benchmarks.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and technically precise comments. These observations identify areas where the manuscript can be strengthened by making implicit technical assumptions explicit. We respond to each major comment below and indicate the corresponding revisions.

read point-by-point responses

  1. Referee: [Framework construction (abstract and §3)] Abstract and framework construction: the driver-conditional risk-neutral measures are obtained by nonlinear filtering of latent structural drivers onto the observable filtration, after which a martingale representation is invoked to equate pricing, hedging, and control. No explicit conditions on the observation process, identifiability of the drivers, or preservation of the predictable representation property are stated; without them the subsequent projection-divergence duality may select a projection that is not the true minimum-distance allocation under the original measure.

    Authors: We agree that the conditions should be stated explicitly. In the revision we will add Assumption 3.1 requiring the observation process to be a diffusion with uniformly elliptic diffusion matrix and Lipschitz coefficients, guaranteeing strong existence and uniqueness of the nonlinear filter. Lemma 3.2 will then verify that the filter preserves measure equivalence to the original risk-neutral measure and that the predictable representation property holds with respect to the innovation process. These additions ensure the subsequent duality is taken with respect to the original measure. revision: yes

  2. Referee: [Projection-divergence duality (abstract and §4)] Projection-divergence duality: the claim that the causal driver span selects the feasible allocation closest to the unconstrained optimum under a convex divergence presupposes that the filtered measure remains equivalent to the original risk-neutral measure and that the driver span is closed in the relevant L2 space. The manuscript must supply the precise statement, the convexity assumption on the divergence, and the verification that the duality is not circular with the filtering step.

    Authors: Theorem 4.1 states the duality precisely: for any convex, lower-semicontinuous divergence D, the L2-projection onto the closed linear span of the causal drivers yields the minimum-distance allocation. Equivalence of the filtered measure is established in Proposition 3.4 before the duality is derived, so the argument is sequential. We will add an explicit convexity assumption in Definition 4.1 and a remark confirming that the driver span is closed in L2 under the maintained ellipticity condition. revision: yes

  3. Referee: [Causal completeness condition (abstract and §5)] Causal completeness condition: the condition identifying when a finite driver span captures systematic premia is load-bearing for the claim that CPCMs generalize classical models. The empirical selection among >300 candidate drivers introduces a free parameter whose effect on the completeness threshold must be quantified; otherwise the reported persistence of premia may be driven by post-selection bias rather than the causal structure.

    Authors: Definition 5.1 requires that the selected driver span equals the space of systematic risk factors under the risk-neutral measure. Driver selection is performed via the PC algorithm with FDR control; the completeness threshold is the numerical rank of the resulting loading matrix. In the revision we will add a sensitivity analysis (new subsection 6.3 and Figure 7) that recomputes Sharpe ratios, turnover, and premia persistence across a grid of FDR levels. This directly quantifies the dependence of the completeness threshold and reported performance on the selection parameter. revision: partial

Circularity Check

0 steps flagged

No significant circularity in the core derivation chain.

full rationale

The paper's central construction begins with the definition of driver-conditional risk-neutral measures via nonlinear filtering on the observable filtration, followed by invocation of a martingale representation theorem to link pricing, hedging, and control. From this foundation it derives the projection-divergence duality and causal completeness condition as consequences within the same information geometry. No quoted equations or steps in the abstract or described framework reduce these results to the inputs by construction, nor do they rely on self-citation load-bearing, fitted parameters renamed as predictions, or ansatzes smuggled via prior work. The empirical selection of drivers and comparison to benchmarks is presented as validation rather than part of the theoretical derivation, leaving the claimed results self-contained against external benchmarks such as the martingale representation property under partial information.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The abstract relies on domain assumptions about the existence and filterability of causal drivers in financial markets and introduces new constructs such as the causal driver span without external verification.

free parameters (1)
  • number of candidate drivers
    More than 300 drivers selected for the U.S. equity panel empirical evaluation.
axioms (1)
  • domain assumption Structural causal drivers exist and are recoverable via nonlinear filtering from partial observations
    Required to construct driver-conditional risk-neutral measures and martingale representation on the observable filtration.
invented entities (1)
  • causal driver span no independent evidence
    purpose: Restriction of portfolios to quantify stability cost via projection-divergence duality
    New construct introduced to select feasible allocations closest to the unconstrained optimum.

pith-pipeline@v0.9.0 · 5731 in / 1568 out tokens · 67412 ms · 2026-05-18T18:10:02.710430+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

76 extracted references · 76 canonical work pages · 1 internal anchor

  1. [1]

    Markowitz, Portfolio selection,The Journal of Finance.7(1), 77–91 (1952)

    H. Markowitz, Portfolio selection,The Journal of Finance.7(1), 77–91 (1952). ISSN 00221082, 15406261. URLhttp://www.jstor.org/stable/ 2975974

    work page 1952

  2. [2]

    Sharpe, Capital asset prices: A theory of market equilibrium under con- ditions of risk,Journal of Finance.19(3), 425–442 (1964)

    W. Sharpe, Capital asset prices: A theory of market equilibrium under con- ditions of risk,Journal of Finance.19(3), 425–442 (1964)

    work page 1964

  3. [3]

    S. A. Ross, The arbitrage theory of capital asset pricing,Journal of Eco- nomic Theory.13(3), 341–360 (1976). URLhttp://www.sciencedirect. com/science/article/pii/0022053176900466

    work page arXiv 1976

  4. [4]

    Black and R

    F. Black and R. Litterman, Global portfolio optimization,Financial Analysts Journal.48(5), 28–43 (1992)

    work page 1992

  5. [5]

    Cambridge University Press (2014)

    R.RebonatoandA.Denev,Portfolio Management under Stress: A Bayesian- Net Approach to Coherent Asset Allocation. Cambridge University Press (2014). ISBN 9781107048119. doi: 10.1017/CBO9781107256736

    work page doi:10.1017/cbo9781107256736 2014

  6. [6]

    Meucci, Fully flexible views: Theory and practice,Risk.21(10), 97–102 (2008)

    A. Meucci, Fully flexible views: Theory and practice,Risk.21(10), 97–102 (2008)

    work page 2008

  7. [7]

    R. C. Merton, An intertemporal capital asset pricing model,Econometrica. 41(5), 867–887 (1973)

    work page 1973

  8. [8]

    Babuška, J

    J. Moody, L. Wu, Y. Liao, and M. Saffell, Performance functions and reinforcement learning for trading systems and portfolios,Journal of Forecasting.17(5-6), 441–470 (1998). doi: https://doi.org/10.1002/(SICI) 1099-131X(1998090)17:5/6<441::AID-FOR707>3.0.CO;2-\#. URLhttps: //onlinelibrary.wiley.com/doi/abs/10.1002/%28SICI%291099-131X% 281998090%2917%3A5/6...

    work page doi:10.1002/(sici 1998

  9. [9]

    M. G. Bellemare, W. Dabney, and M. Rowland,Distributional Reinforcement Learning. MIT Press (2023).http://www.distributional-rl.org

    work page 2023

  10. [10]

    Buehler, L

    H. Buehler, L. Gonon, J. Teichmann, and B. Wood, Deep hedging,Quantita- November 14, 2025 1:40 ws-rv9x6 Book Title ws-rv9x6 page 75 Causal PDE–Control for Adaptive Portfolio Optimization under Partial Information75 tive Finance.19(8), 1271–1291 (2019). doi: 10.1080/14697688.2019.1571683. URLhttps://doi.org/10.1080/14697688.2019.1571683

    work page doi:10.1080/14697688.2019.1571683 2025

  11. [11]

    M.DixonandI.Halperin, G-learnerandgirl: Goalbasedwealthmanagement with reinforcement learning,SSRN Electronic Journal(01, 2020). doi: 10. 2139/ssrn.3543852

    work page 2020

  12. [12]

    Zhang, C

    J. Zhang, C. Wan, M. Chen, and H. Liu, An efficient reinforcement learn- ing approach for goal-based wealth management,Expert Systems with Ap- plications.237, 121578 (2024). ISSN 0957-4174. doi: https://doi.org/10. 1016/j.eswa.2023.121578. URLhttps://www.sciencedirect.com/science/ article/pii/S0957417423020808

    work page arXiv 2024

  13. [13]

    Correlation

    D. DUFFIE, A. ECKNER, G. HOREL, and L. SAITA, Frailty correlated default,The Journal of Finance.64(5), 2089–2123 (2009). doi: https: //doi.org/10.1111/j.1540-6261.2009.01495.x. URLhttps://onlinelibrary. wiley.com/doi/abs/10.1111/j.1540-6261.2009.01495.x

    work page doi:10.1111/j.1540-6261.2009.01495.x 2089

  14. [14]

    Pareek and S

    S. Pareek and S. Ghosh, Semiparametric dynamic copula approach to portfo- lio selection,arXiv preprint arXiv:2504.12266(2025). URLhttps://arxiv. org/abs/2504.12266

    work page arXiv 2025

  15. [15]

    Ito and T

    K. Ito and T. Yoshiba, Dynamic asymmetric tail dependence struc- ture among multi-asset classes for portfolio management: Dynamic skew-t copula approach,International Review of Economics & Fi- nance.97, 103724 (2025). ISSN 1059-0560. doi: https://doi.org/10. 1016/j.iref.2024.103724. URLhttps://www.sciencedirect.com/science/ article/pii/S1059056024007160

    work page arXiv 2025

  16. [16]

    Øksendal and A

    B. Øksendal and A. Sulem, Portfolio optimization under model uncer- tainty and bsde games,Quantitative Finance.11(11), 1665–1674 (2011). doi: 10.1080/14697688.2011.615219. URLhttps://doi.org/10.1080/14697688. 2011.615219

    work page doi:10.1080/14697688.2011.615219 2011

  17. [17]

    Karatzas and R

    I. Karatzas and R. Fernholz. Stochastic portfolio theory: an overview. In eds. A. Bensoussan and Q. Zhang,Special Volume: Mathematical Modeling and Numerical Methods in Finance, vol. 15,Handbook of Numerical Analysis, pp. 89–167. Elsevier (2009). doi: 10.1016/S1570-8659(08)00003-3. URLhttps: //www.sciencedirect.com/science/article/pii/S1570865908000033

    work page doi:10.1016/s1570-8659(08)00003-3 2009

  18. [18]

    J.-W. Gu, S. Si, and H. Zheng, Constrained utility deviation-risk optimiza- tion and time-consistent hjb equation,SIAM Journal on Control and Op- timization.58(2), 866–894 (2020). doi: 10.1137/19M1256014. URLhttps: //doi.org/10.1137/19M1256014

    work page doi:10.1137/19m1256014 2020

  19. [19]

    A. Rodriguez Dominguez, Portfolio optimization based on neural networks sensitivities from assets dynamics respect common drivers,Machine Learn- ing with Applications.11, 100447 (2023). ISSN 2666-8270. doi: https: //doi.org/10.1016/j.mlwa.2022.100447. URLhttps://www.sciencedirect. com/science/article/pii/S2666827022001220

    work page doi:10.1016/j.mlwa.2022.100447 2023

  20. [20]

    Lintner, The valuation of risk assets and the selection of risky investments in stock portfolios and capital budgets,The Review of Economics and Statis- tics.47(1), 13–37 (1965)

    J. Lintner, The valuation of risk assets and the selection of risky investments in stock portfolios and capital budgets,The Review of Economics and Statis- tics.47(1), 13–37 (1965). URLhttp://www.jstor.org/stable/1924119

    work page arXiv 1965

  21. [21]

    E. F. Fama and K. R. French, The cross-section of expected stock returns, The Journal of Finance.47(2), 427–465 (1992). November 14, 2025 1:40 ws-rv9x6 Book Title ws-rv9x6 page 76 76Alejandro Rodriguez Dominguez

    work page 1992

  22. [22]

    E. F. Fama and K. R. French, Common risk factors in the returns on stocks and bonds,Journal of Financial Economics.33(1), 3–56 (1993)

    work page 1993

  23. [23]

    DeMiguel, L

    V. DeMiguel, L. Garlappi, and R. Uppal, Optimal versus naive diversifica- tion: How inefficient is the 1/n portfolio strategy?,The Review of Financial Studies.22(5), 1915–1953 (2009)

    work page 1915

  24. [24]

    P. N. Kolm and G. Ritter, Factor investing with black–litterman–bayes: In- corporating factor views and priors in portfolio construction,Journal of Port- folio Management.47(2), 113–126 (2021). doi: 10.3905/jpm.2020.1.196

    work page doi:10.3905/jpm.2020.1.196 2021

  25. [25]

    L. P. Hansen and T. J. Sargent,Robustness, course book edn. Princeton University Press, Princeton, NJ (2011). ISBN 9781400829385

    work page 2011

  26. [26]

    Bellman, Dynamic programming,Science.153(3731), 34–37 (1966)

    R. Bellman, Dynamic programming,Science.153(3731), 34–37 (1966). doi: 10.1126/science.153.3731.34. URLhttps://www.science.org/doi/abs/10. 1126/science.153.3731.34

    work page doi:10.1126/science.153.3731.34 1966

  27. [27]

    W. H. Fleming and H. M. Soner,Controlled Markov Processes and Viscosity Solutions, 2nd edn. Springer (2006)

    work page 2006

  28. [28]

    Yong and X

    J. Yong and X. Y. Zhou,Stochastic Controls: Hamiltonian Systems and HJB Equations. vol. 43,Stochastic Modelling and Applied Probability, Springer, New York (1999). doi: 10.1007/b97848

    work page doi:10.1007/b97848 1999

  29. [29]

    Cvitanić and I

    J. Cvitanić and I. Karatzas, Convex duality in constrained portfolio op- timization,The Annals of Applied Probability.2(4), 767–818 (1992). URL http://www.jstor.org/stable/2959666

    work page arXiv 1992

  30. [30]

    Trimborn, L

    T. Trimborn, L. Pareschi, and M. Frank, Portfolio optimization and model predictive control: A kinetic approach,Discrete and Continuous Dynamical Systems - B.24(11), 6209–6238 (2019). doi: 10.3934/dcdsb.2019136. URLhttps://www.aimsciences.org/article/id/ df0faf53-7301-48e9-b088-37f657c1111d

    work page doi:10.3934/dcdsb.2019136 2019

  31. [31]

    H. J. Kushner and P. G. Dupuis,Numerical Methods for Stochastic Con- trol Problems in Continuous Time, 2nd edn. vol. 24,Applications of Math- ematics, Springer, New York (2001). ISBN 9780387952598. doi: 10.1007/ 978-1-4613-0009-0

    work page 2001

  32. [32]

    Delage and Y

    E. Delage and Y. Ye, Distributionally robust optimization under moment uncertainty with application to data-driven problems,Operations Research. 58, 595–612 (06, 2010). doi: 10.1287/opre.1090.0741

    work page doi:10.1287/opre.1090.0741 2010

  33. [33]

    P. M. Esfahani and D. Kuhn, Data-driven distributionally robust optimiza- tion using the wasserstein metric: Performance guarantees and tractable re- formulations,Mathematical Programming.171(1), 115–166 (2018)

    work page 2018

  34. [34]

    Nakayama and T

    Y. Nakayama and T. Sawaki, Causal inference on investment constraints and non-stationarity in dynamic portfolio optimization through reinforcement learning,arXiv preprint.abs/2311.04946(2023). URLhttps://arxiv. org/abs/2311.04946

    work page arXiv 2023

  35. [35]

    Ruf and W

    J. Ruf and W. Wang, Hedging with linear regressions and neural networks, Journal of Business & Economic Statistics.40(4), 1442–1454 (2022). doi: 10. 1080/07350015.2021.1931241. URLhttps://doi.org/10.1080/07350015. 2021.1931241

    work page doi:10.1080/07350015 2022

  36. [36]

    Fu and A

    W. Fu and A. Hirsa. Solving barrier options under stochastic volatility using deep learning. (2022). URLhttps://api.semanticscholar.org/CorpusID: November 14, 2025 1:40 ws-rv9x6 Book Title ws-rv9x6 page 77 Causal PDE–Control for Adaptive Portfolio Optimization under Partial Information77 250243619

    work page 2022

  37. [37]

    Bisht and A

    K. Bisht and A. Kumar, A portfolio construction model based on sec- tor analysis using dempster-shafer evidence theory and granger causal net- work: An application to national stock exchange of india,Expert Sys- tems with Applications.215, 119434 (2023). ISSN 0957-4174. doi: 10. 1016/j.eswa.2022.119434. URLhttps://www.sciencedirect.com/science/ article/pi...

    work page arXiv 2023

  38. [38]

    J. Han, A. Jentzen, and W. E, Solving high-dimensional partial differen- tial equations using deep learning,Proceedings of the National Academy of Sciences.115(34), 8505–8510 (2018)

    work page 2018

  39. [39]

    Sirignano and K

    J. Sirignano and K. Spiliopoulos, Dgm: A deep learning algorithm for solving partial differential equations,Journal of Computational Physics.375, 1339– 1364 (2018). doi: 10.1016/j.jcp.2018.08.029

    work page doi:10.1016/j.jcp.2018.08.029 2018

  40. [40]

    Raissi, P

    M. Raissi, P. Perdikaris, and G. E. Karniadakis, Physics-informed neural net- works: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,Journal of Computational Physics.378, 686–707 (2019)

    work page 2019

  41. [41]

    Noguer i Alonso and J

    M. Noguer i Alonso and J. Antolín Camarena, Physics-informed neural net- works (pinns) in finance,SSRN Electronic Journal(October, 2023). doi: 10.2139/ssrn.4598180. URLhttps://ssrn.com/abstract=4598180

    work page doi:10.2139/ssrn.4598180 2023

  42. [42]

    Pearl,Causality: Models, Reasoning and Inference, 2nd edn

    J. Pearl,Causality: Models, Reasoning and Inference, 2nd edn. Cambridge University Press (2009)

    work page 2009

  43. [43]

    Peters, D

    J. Peters, D. Janzing, and B. Schölkopf,Elements of Causal Inference: Foun- dations and Learning Algorithms. MIT Press (2017)

    work page 2017

  44. [44]

    Chernozhukov, C

    V. Chernozhukov, C. Cinelli, W. Newey, A. Sharma, and V. Syrgkanis. Long story short: Omitted variable bias in causal machine learning. Working Paper 30302, National Bureau of Economic Research (July, 2022). URLhttp:// www.nber.org/papers/w30302

    work page 2022

  45. [45]

    Bensoussan,Stochastic Control of Partially Observable Systems

    A. Bensoussan,Stochastic Control of Partially Observable Systems. Cam- bridge Series in Statistical and Probabilistic Mathematics, Cambridge Uni- versity Press, Cambridge (1992). ISBN 9780521415343. doi: 10.1017/ CBO9780511574767

    work page 1992

  46. [46]

    A.BainandD.Crisan,Fundamentals of Stochastic Filtering.vol.60,Stochas- tic Modelling and Applied Probability, Springer (2009)

    work page 2009

  47. [47]

    Invariant Risk Minimization

    M. Arjovsky, L. Bottou, I. Gulrajani, and D. Lopez-Paz, Invariant risk minimization,ArXiv.abs/1907.02893(2019). URLhttps://api. semanticscholar.org/CorpusID:195820364

    work page internal anchor Pith review Pith/arXiv arXiv 1907

  48. [48]

    Pham,Continuous-time Stochastic Control and Optimization with Finan- cial Applications

    H. Pham,Continuous-time Stochastic Control and Optimization with Finan- cial Applications. Springer (2009)

    work page 2009

  49. [49]

    Carmona and F

    R. Carmona and F. Delarue,Probabilistic Theory of Mean Field Games with Applications, Volume I: Mean Field FBSDEs, Control, and Games. vol. 84, Probability Theory and Stochastic Modelling, Springer, Cham (2018). ISBN 978-3-319-58919-0. doi: 10.1007/978-3-319-58920-6

    work page doi:10.1007/978-3-319-58920-6 2018

  50. [50]

    L. C. Evans,Partial Differential Equations, 2 edn. vol. 19,Graduate Stud- ies in Mathematics, American Mathematical Society, Providence, RI (2010). ISBN 978-0-8218-4974-3. November 14, 2025 1:40 ws-rv9x6 Book Title ws-rv9x6 page 78 78Alejandro Rodriguez Dominguez

    work page 2010

  51. [51]

    Jacod and A

    J. Jacod and A. N. Shiryaev,Limit Theorems for Stochastic Processes, 2 edn. vol. 288,Grundlehren der mathematischen Wissenschaften, Springer (2003). ISBN 978-3-540-43932-5. doi: 10.1007/978-3-662-05265-5

    work page doi:10.1007/978-3-662-05265-5 2003

  52. [52]

    P. E. Protter,Stochastic Integration and Differential Equations, 2nd edn. vol. 21,Stochastic Modelling and Applied Probability, Springer, Berlin (2005). ISBN 9783540003137. doi: 10.1007/978-3-662-10061-5

    work page doi:10.1007/978-3-662-10061-5 2005

  53. [53]

    Kallianpur,Stochastic Filtering Theory

    G. Kallianpur,Stochastic Filtering Theory. vol. 13,Applications of Mathe- matics, Springer (1980). doi: 10.1007/978-1-4612-6003-1

    work page doi:10.1007/978-1-4612-6003-1 1980

  54. [54]

    R. S. Liptser and A. N. Shiryaev,Statistics of Random Processes I: General Theory, 1 edn. Applications of Mathematics, Springer, Berlin, Heidelberg (1977)

    work page 1977

  55. [55]

    2013 , isbn =

    J. Xiong,An Introduction to Stochastic Filtering Theory. Oxford University Press(04, 2008).ISBN9780199219704.doi: 10.1093/oso/9780199219704.001

    work page doi:10.1093/oso/9780199219704.001 2008

  56. [56]

    URLhttps://doi.org/10.1093/oso/9780199219704.001.0001

    work page doi:10.1093/oso/9780199219704.001.0001

  57. [57]

    H.Reichenbach,The Direction of Time.UniversityofCaliforniaPress, Berke- ley (1956)

    work page 1956

  58. [58]

    Inada, On a two-sector model of economic growth: Comments and a generalization1,The Review of Economic Studies.30(2), 119–127 (06, 1963)

    K.-i. Inada, On a two-sector model of economic growth: Comments and a generalization1,The Review of Economic Studies.30(2), 119–127 (06, 1963). ISSN 0034-6527. doi: 10.2307/2295809. URLhttps://doi.org/10. 2307/2295809

    work page doi:10.2307/2295809 1963

  59. [59]

    M. G. Crandall, H. Ishii, and P.-L. Lions, User’s guide to viscosity solutions of second order partial differential equations,Bull. Amer. Math. Soc.27, 1–67 (1992)

    work page 1992

  60. [60]

    Kalman, A new approach to linear filtering and prediction problems,Jour- nal of Basic Engineering.82, 35–45 (1960)

    R. Kalman, A new approach to linear filtering and prediction problems,Jour- nal of Basic Engineering.82, 35–45 (1960)

    work page 1960

  61. [61]

    Jazwinski,Stochastic Processes and Filtering Theory

    A. Jazwinski,Stochastic Processes and Filtering Theory. Academic Press (1970)

    work page 1970

  62. [62]

    Doucet, N

    A. Doucet, N. de Freitas, and N. Gordon. Sequential monte carlo methods in practice. InSequential Monte Carlo Methods in Practice. Springer (2001)

    work page 2001

  63. [63]

    Doucet and A

    A. Doucet and A. Johansen, A tutorial on particle filtering and smoothing: Fifteen years later,Handbook of Nonlinear Filtering.12(01, 2009)

    work page 2009

  64. [64]

    Cappé, E

    O. Cappé, E. Moulines, and T. Rydén,Inference in Hidden Markov Models. Springer (2005)

    work page 2005

  65. [65]

    Shreve,Stochastic Calculus for Finance II: Continuous-Time Models

    S. Shreve,Stochastic Calculus for Finance II: Continuous-Time Models. Springer (2004)

    work page 2004

  66. [66]

    Brigo and F

    D. Brigo and F. Mercurio,Interest Rate Models—Theory and Practice, 2 edn. Springer (2006)

    work page 2006

  67. [67]

    Björk,Arbitrage Theory in Continuous Time, 3 edn

    T. Björk,Arbitrage Theory in Continuous Time, 3 edn. Oxford University Press (2009)

    work page 2009

  68. [68]

    Merlevède, M

    F. Merlevède, M. Peligrad, and E. Rio, Bernstein inequality and moderate deviations under strong mixing conditions,The Annals of Probability.37(6), 2059–2143 (2009)

    work page 2059

  69. [69]

    Rio,Théorie asymptotique des processus aléatoires faiblement dépendants

    E. Rio,Théorie asymptotique des processus aléatoires faiblement dépendants. Springer (2000)

    work page 2000

  70. [70]

    K.-i. Yoshihara, Limiting behavior of u-statistics for stationary, absolutely regular processes,Zeitschrift für Wahrscheinlichkeitstheorie und Verwandte November 14, 2025 1:40 ws-rv9x6 Book Title ws-rv9x6 page 79 Causal PDE–Control for Adaptive Portfolio Optimization under Partial Information79 Gebiete.35, 237–252 (1976)

    work page 2025

  71. [71]

    M. A. Arcones and B. Yu, Central limit theorems for empirical and u- processes of stationary mixing sequences,Stochastic Processes and their Ap- plications.54(2), 231–253 (1994)

    work page 1994

  72. [72]

    Friedman,Partial Differential Equations of Parabolic Type

    A. Friedman,Partial Differential Equations of Parabolic Type. Courier Dover Publications (2008). ISBN 9780486462905

    work page 2008

  73. [73]

    Optimal Transport for Applied Mathematicians: Calculus of Variations, PDEs, and Modeling , year =

    F. Santambrogio,Optimal Transport for Applied Mathematicians. vol. 87, Progress in Nonlinear Differential Equa- tions and Their Applications, Birkhäuser (2015). ISBN 978-3-319-20827-5. doi: 10.1007/978-3-319-20828-2

    work page doi:10.1007/978-3-319-20828-2 2015

  74. [74]

    Bolley and C

    F. Bolley and C. Villani, Weighted csiszár–kullback–pinsker inequalities and applications to transportation inequalities,Annales de la Faculté des sciences de Toulouse, 6e série.14(3), 331–352 (2005). URLhttp://www.numdam.org/ item?id=AFST_2005_6_14_3_331_0

    work page 2005

  75. [75]

    Ambrosio, N

    L. Ambrosio, N. Gigli, and G. Savaré,Gradient Flows: In Metric Spaces and in the Space of Probability Measures, 2nd edn. Lectures in Mathemat- ics. ETH Zürich, Birkhäuser (2008). ISBN 9783764387228. doi: 10.1007/ 978-3-7643-8722-8. URLhttps://doi.org/10.1007/978-3-7643-8722-8

    work page doi:10.1007/978-3-7643-8722-8 2008

  76. [76]

    Villani , Optimal Transport, vol

    C. Villani,Optimal Transport: Old and New. vol. 338,Grundlehren der math- ematischen Wissenschaften, Springer (2008). ISBN 978-3-540-71049-3. doi: 10.1007/978-3-540-71050-9

    work page doi:10.1007/978-3-540-71050-9 2008