Pith. sign in

REVIEW 2 major objections 7 minor 92 references

The paper establishes an invertible coordinate change Ψ on the tensor algebra under which the expected signature of an augmented path (A,X) becomes a deterministic simplex integral of X-correlators, so expected signatures can be computed by

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review

2026-08-03 05:05 UTC pith:PCPXM6ZJ

load-bearing objection The Ψ-transform is a real and useful coordinate change, but the paper's advertised law-determinacy theorem rests on a misreading of Petersen's theorem and needs an extra assumption. the 2 major comments →

arxiv 2607.29534 v1 pith:PCPXM6ZJ submitted 2026-07-31 math.PR

Expected signatures via partial integration, coordinate change and symmetrization

Paul P. Hager , Luca Pelizzari This is my paper
classification math.PR MSC 60L7060G22
keywords path signatureexpected signaturecoordinate transformationpartial integrationrough pathsmoment problemfractional Brownian motionstochastic control
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper proposes a way to compute expected signatures of processes of the form Y=(A,X), where A is a regular deterministic augmentation component and X is a possibly rough stochastic component. Its central claim is that there is an explicit invertible linear change of coordinates Ψ on the tensor algebra such that, in Ψ-coordinates, the signature of Y contains no mixed iterated integrals against dX: every mixed word becomes an iterated integral against dA of ordinary signature coordinates of X. Taking expectations then commutes with the deterministic integrals, so the expected signature is a simplex integral of correlators of X. For Gaussian or polynomial X these correlators are available in closed form, so expected signatures can be obtained by quadrature rather than by averaging signatures of discretized sample paths, which for fractional Brownian motion suffers a slow Δt^(2H) bias. The paper also proves law-determinacy for the partially symmetrized expected signature under a moment-determinacy condition on the one-dimensional marginals, and demonstrates the method on a signature-based stochastic control problem.

Core claim

Theorem 3.6 (with Prop 3.2 and Prop 4.1) states that for every word w=w_0 i_1 w_1 ... i_k w_k, the Ψ-coordinate of the signature of Y=(A,X) equals ⟨w_k, Sig(X)⟩ times an iterated integral over the simplex of products ⟨w_{j-1}, Sig(X)⟩ against dA^{i_j}. Consequently the expected signature is a deterministic simplex integral whose integrand is a correlator of X. The partially symmetrized version replaces the signature factors of X by normalized multivariate increments, giving explicit correlators; for Gaussian processes these reduce to sums of products of covariance functions via the classical Wick formula. Under Assumption 2.3 (e.g., A contains time) and moment-determinate one-dimensional mar

What carries the argument

The central object is the graded automorphism Ψ (and its partial-symmetrization descendant bΨ), constructed recursively by partial integration and the shuffle product. On a word w, Ψ replaces mixed blocks by shuffling the terminal X-block against the recursively transformed prefix, so that dX never appears as an integration variable; the inverse Ψ^{-1} is defined by the companion recursion. The induced pairing ⟨α,a⟩_Ψ := ⟨α, Ψ(a)⟩ makes the signature coordinates explicit, and the dual product • (resp. b• after quotienting by A''-commutators) gives the algebra structure in which the transformed signature is group-like. This machinery turns expected-signature evaluation into deterministic simp

Load-bearing premise

The law-determinacy theorems assume that equality of all joint moments, together with moment-determinacy of the coordinate marginals of X, forces equality of the finite-dimensional laws; this is only true if every one-dimensional projection of X is moment-determinate, not just the coordinate axes, and without that stronger condition there are distinct laws with identical moments.

What would settle it

Take A_t = t and let X be a random linear path X_t = (Z,W)t, where (Z,W) under P and Q are two distinct laws on R^2 with identical all joint moments and standard normal coordinate marginals (such non-Gaussian laws sharing every moment with a bivariate normal are known to exist). Since the partially symmetrized expected signature of (A,X) encodes exactly the joint moments of (Z,W), the two expected signatures coincide while P and Q differ, contradicting the stated equivalence if the moment-determinacy assumption is read as only coordinate-marginal.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X LinkedIn Reddit HN

If this is right

  • Any expected signature of an augmented Gaussian or polynomial process can be evaluated by quadrature once its correlators are known, bypassing Monte Carlo on discretized paths.
  • The mixed grading reduces tensor dimension: keeping the X-degree at 1 while taking the A-degree up to 9 makes the control problem tractable, while the full (21,21) truncation would contain over two trillion entries.
  • The naive piecewise-linear Monte Carlo estimator for the expected signature has weak bias of order Δt^(2H) for fractional Brownian motion; the Ψ-coordinate method avoids this bias.
  • Partially symmetrized expected signatures are law-determining when the augmentation is rich and the coordinate marginals of X are moment-determinate, and linear functionals in Ψ-coordinates provide universal approximation classes.
  • The coordinate representation canonically produces joint rough path lifts when A is a Young-integrable or semimartingale component, unifying several known lift constructions.

Where Pith is reading between the lines

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

  • One testable extension is to non-independent stochastic augmentations A, where the correlators of X no longer factor from the augmentation; the paper leaves this open, and a useful formula would need joint correlators of (A,X).
  • The partial-symmetrization quotient suggests an algebraic completion of partial rough path spaces used in rough volatility models; building numerical schemes directly on the completed bΨ-coordinates could be a natural next step.
  • For fractional Brownian motion, the closed-form beta-function expressions given for mixed degree (n_A,2) could be extended recursively to higher X-degrees, potentially yielding fully explicit expected signatures for that model.
  • The law-determinacy theorem is stated under coordinate-marginal moment-determinacy; whether the conclusion survives without strengthening to all one-dimensional projections is a question worth resolving, as standard multidimensional moment counterexamples live exactly at that point.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 7 minor

Summary. The paper introduces an invertible graded linear transformation Ψ on the tensor algebra over a split alphabet A=A'∪A'', with the property that in Ψ-coordinates the signature of (A,X) is expressed as iterated integrals against the regular component A of terms built from the signature of X (Props. 3.2, 3.4, Thm. 3.6). It develops a partially symmetrized version bΨ, identifies the associated dual products and Hopf-algebra structures, and uses the representation to compute expected signatures of augmented processes: for deterministic A, expected Ψ-coordinates are simplex integrals of correlators of X (Prop. 4.1). Applications include closed-form correlators for Gaussian and polynomial processes, a quadrature-based method for fractional Brownian motion expected signatures, a signature-based control example, and a law-determinacy theorem (Thm. 1.5, Props. 4.2, 4.6).

Significance. The Ψ-transform construction is explicit, self-contained, and, as far as I can verify, correct: invertibility is proved from the shuffle identity and the spanning property of truncated signatures, with no fitted parameters. The resulting representation cleanly separates probabilistic correlators from deterministic simplex integration, and the numerical experiments for fBm, supported by a companion implementation, give a credible and useful computational recipe. If the law-determinacy claims are repaired, this would be a solid contribution to expected-signature computation and to signature-based methods for heterogeneous paths.

major comments (2)
  1. [§4.1, Prop. 4.2 and Thm. 1.5] The decisive step "assumption (54) implies … the finite-dimensional law is uniquely determined by its multivariate moments" misreads Petersen's theorem [84, Thm. 3]. That theorem requires moment-determinacy of every one-dimensional projection ℓ·X of the finite-dimensional vector, not merely of the coordinate marginals X^i_t. Condition (54) only gives determinacy along the coordinate projections e_i. There are classical multidimensional moment-problem examples with identical joint moments, identical determinate coordinate marginals, but different joint laws. Equality of all correlators bC_P=bC_Q is exactly equality of all mixed moments at all time tuples, so it does not imply P=Q. This affects Theorem 1.5, Corollary 4.3, and Proposition 4.6. The theorem can be repaired by strengthening (54) to moment-determinacy of all one-dimensional projections (or of the relevant finite-dimensional law
  2. [§4.2, Prop. 4.6] The Brownian augmentation case inherits the same gap. Lemma 4.5 shows that Ψ-coordinates of bμ recover the same integrated correlators as in the deterministic-time case, but the final inference from equality of all mixed moments to P=Q again invokes the same invalid Petersen step. The same strengthened moment-determinacy assumption on projections is needed; otherwise the conclusion does not follow.
minor comments (7)
  1. [Title/Abstract] "SIGNA TURES" in the title contains a spurious space.
  2. [§1, after Cor. 1.4] "This gain in tractability … does not have come at the cost" should read "does not come at the cost".
  3. [Proof of Thm. 3.6] There is a doubled angle bracket: "⟨w,Ψ^{-1}(Ψ(a))⟩⟩".
  4. [Proof of Thm. 3.10] "we first the following" is missing a verb; it should be "we first prove the following".
  5. [Eq. (66)] The notation "•2•" is unclear. If it denotes the •-square of an element, please write it as e.g. "α^{•2}" or "α•α" and define it.
  6. [Appendix A, Lemma A.1] The statement that the rate is "sharp" because there are words with error o(Δt^{2H}) is not consistent: o(Δt^{2H}) would show the error is smaller than the advertised order, not that the rate cannot be improved. If the intended claim is that the error for w=221 is of exact order Δt^{2H}, the argument needs correction, since the standard trapezoidal-rule estimate is not uniform as the interval endpoint tends to 0.
  7. [Reference [84]] The author field "LC690537 Petersen" appears corrupted; the reference should be to the actual author and paper (Petersen, Math. Scand., 1982).

Circularity Check

0 steps flagged

No significant circularity: the Ψ-transform derivation is self-contained and the expected-signature formulas follow from definitions plus external results.

full rationale

The central derivation chain is not circular. The coordinate transformation Ψ is constructed explicitly in Proposition 3.2 by a recursion based on integration by parts and the shuffle identity, and its inverse is constructed in Proposition 3.4. Invertibility of Ψ is proved in Theorem 3.6 using the independent spanning result [42, Lemma 5], which is an external citation, not a self-citation. The expected-signature representation in Proposition 4.1 is obtained by linearity of expectation applied to the pathwise representation, with no parameter fitted to data and no target quantity renamed as a prediction. The Gaussian and polynomial correlator formulas in Examples 5 and 6 use external results (Isserlis' theorem and [15, Theorem 4.5]) and are not used to define the objects they claim to compute. The paper's self-citations ([49,50] on signature cumulants, [62] on implementation of truncated signatures) are contextual or computational and are not load-bearing for the main analytical claims. The moment-determinacy argument in Section 4.1 cites Petersen's theorem [84, Theorem 3], an external mathematical result; whether condition (54) is sufficient as used is a substantive correctness question, but it is not circularity because the paper does not define the conclusion in terms of the citation or fit the result to itself. Overall, the derivation of the expected-signature formulas is self-contained against external benchmarks, so the appropriate circularity score is 0.

Axiom & Free-Parameter Ledger

0 free parameters · 7 axioms · 0 invented entities

The central claims rest on standard rough-path/signature theorems and a few domain assumptions. No free parameters are fitted anywhere in the derivation. The fragile premise is the moment-determinacy bridge in Section 4.1: coordinate-marginal moment-determinacy alone is insufficient for finite-dimensional law determinacy in multiple dimensions.

axioms (7)
  • standard math Signature satisfies Chen's relation and the shuffle identity
    Used throughout Section 3 to derive and invert Ψ-coordinates; standard property of geometric rough paths.
  • standard math Lyons extension theorem gives a unique full signature for a geometric p-rough path
    Needed in Corollary 3.14 to define Sig(X) canonically from the rough path enhancement.
  • standard math Truncated signatures of smooth paths span the truncated tensor algebra ([42, Lemma 5])
    Used in Theorem 3.6 to lift coordinate identities from signature elements to all tensor algebra elements.
  • domain assumption Coordinate-marginal moment-determinacy plus equality of mixed moments implies joint law determinacy
    Assumed in Section 4.1 through [84, Theorem 3]; the cited theorem requires all one-dimensional projections, so this is stronger than justified and is the weakest premise.
  • domain assumption Assumption 2.3: derivatives of A-signature coordinates form a total family in L^2
    Used in Theorem 3.10 injectivity and Prop 4.2 to recover correlators from expected signatures; holds for a time component.
  • domain assumption One-dimensional marginals of X_t^i are moment-determinate and all expected signatures/correlators are well defined
    Stated in Theorem 1.5 and Prop 4.2 as sufficient conditions for P=Q.
  • domain assumption For the Brownian augmentation, A and X are independent and Stratonovich/Itô conversion yields the factor 2^{-n}
    Used in Lemma 4.5 to prove Prop 4.6.

reviewed 2026-08-03 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Expected signatures via partial integration, coordinate change and symmetrization." pith.science (2026). https://pith.science/paper/PCPXM6ZJ

@misc{pith2026260729534,
  author       = {Pith},
  title        = {Pith review of: Expected signatures via partial integration, coordinate change and symmetrization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PCPXM6ZJ}},
  note         = {Machine review of arXiv:2607.29534}
}
Share X LinkedIn Reddit HN
read the original abstract

We study signature transformations of heterogeneous paths $Y=(A,X)$ whose components may differ in regularity and probabilistic structure. We introduce an invertible change of coordinates $\Psi$ such that the transformed signature $\Psi\circ\mathrm{Sig}$ eliminates mixed integration against the irregular component $X$ and admits a representation in terms of signature coordinates of $X$ and iterated integration against the regular component $A$. In addition, we exploit this representation to further represent partially symmetrized signatures. Our main application concerns new expected signature formulas for processes with deterministic augmentation. On the analytical side, these formulas are leveraged to study moment problems. On the numerical side, they enable accurate computation of expected signatures, thereby overcoming typical computational bottlenecks in applications. We illustrate these advantages in a signature-based stochastic control problem driven by fractional Brownian motion.

Figures

Figures reproduced from arXiv: 2607.29534 by Luca Pelizzari, Paul P. Hager.

Figure 1
Figure 1. Figure 1: Runtime versus maximal absolute error for the computation of the trun￾cated expected signature of the time-augmented fractional Brownian motion. Solid lines with circular markers correspond to the conventional Monte Carlo estimator based on piecewise-linear samples, while dashed lines with star markers correspond to the correlator-based Ψ transform method using deterministic quadrature. Colors indicate the… view at source ↗
Figure 2
Figure 2. Figure 2: Left: Relative approximation errors of the mixed-degree controls (NA, 1) with respect to the benchmark, across Hurst parameters H and time degrees NA. Right: Minimal tracking costs obtained from expected signatures in Ψ-coordinates for selected mixed degrees (NA, 1), compared with the benchmark. Recall that the corresponding mixed degree in the optimization problem is (2NA + 3, 2); see (66). Parameters: Y0… view at source ↗
Figure 3
Figure 3. Figure 3: Weak error of the conventional Monte Carlo estimator for the expected signature of the time-augmented fractional Brownian motion. Solid lines correspond to the Monte Carlo errors, dashed lines show the reference rates N −2H [PITH_FULL_IMAGE:figures/full_fig_p038_3.png] view at source ↗

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

92 extracted references · 5 linked inside Pith

  1. [1]

    Stochastic control with signatures via riccati equations on the tensor algebra, 2026

    Eduardo Abi Jaber, Elie Attal, and Dimitri Sotnikov. Stochastic control with signatures via riccati equations on the tensor algebra, 2026

    2026

  2. [2]

    Polynomial Volterra processes.Electronic Journal of Probability, 29:1–37, 2024

    Eduardo Abi Jaber, Christa Cuchiero, Luca Pelizzari, Sergio Pulido, and Sara Svaluto-Ferro. Polynomial Volterra processes.Electronic Journal of Probability, 29:1–37, 2024

    2024

  3. [3]

    Signature volatility models: Pricing and hedging with fourier.SIAM Journal on Financial Mathematics, 16(2):606–642, 2025

    Eduardo Abi Jaber and Louis-Amand G´ erard. Signature volatility models: Pricing and hedging with fourier.SIAM Journal on Financial Mathematics, 16(2):606–642, 2025

    2025

  4. [4]

    Signature approach for pricing and hedging path- dependent options with frictions, 2025

    Eduardo Abi Jaber, Donatien Hainaut, and Edouard Motte. Signature approach for pricing and hedging path- dependent options with frictions, 2025

    2025

  5. [5]

    Signature-based validation of real-world economic scenarios.ASTIN Bulletin, 54(2):410–440, 2024

    Herv´ e Andr` es, Alexandre Boumezoued, and Benjamin Jourdain. Signature-based validation of real-world economic scenarios.ASTIN Bulletin, 54(2):410–440, 2024

    2024

  6. [6]

    Solving linear-quadratic stochastic control problems with signatures.arXiv preprint arXiv:2602.23473, 2026

    Alif Aqsha, Peter Bank, and Leandro S´ anchez-Betancourt. Solving linear-quadratic stochastic control problems with signatures.arXiv preprint arXiv:2602.23473, 2026

    arXiv 2026

  7. [7]

    Stochastic control with signatures.arXiv preprint arXiv:2406.01585, 2024

    Peter Bank, Christian Bayer, Paul P Hager, S Riedel, and T Nauen. Stochastic control with signatures.arXiv preprint arXiv:2406.01585, 2024

    Pith/arXiv arXiv 2024

  8. [8]

    Hedging with temporary price impact.Mathematics and financial economics, 11(2):215–239, 2017

    Peter Bank, H Mete Soner, and Moritz Voß. Hedging with temporary price impact.Mathematics and financial economics, 11(2):215–239, 2017

    2017

  9. [9]

    Operators associated with a stochastic differential equation driven by fractional brownian motions.Stochastic Processes and Their Applications, 117(5):550–574, 2007

    Fabrice Baudoin and Laure Coutin. Operators associated with a stochastic differential equation driven by fractional brownian motions.Stochastic Processes and Their Applications, 117(5):550–574, 2007. 34 PAUL P. HAGER AND LUCA PELIZZARI

    2007

  10. [10]

    Christian Bayer and Peter K. Friz. Cubature on Wiener space: pathwise convergence.Appl. Math. Optim., 67(2):261–278, 2013

    2013

  11. [11]

    Rough volatility

    Christian Bayer, Peter K Friz, Masaaki Fukasawa, Jim Gatheral, Antoine Jacquier, and Mathieu Rosenbaum. Rough volatility. SIAM, 2023

    2023

  12. [12]

    A regularity structure for rough volatility.Mathematical Finance, 30(3):782–832, 2020

    Christian Bayer, Peter K Friz, Paul Gassiat, Jorg Martin, and Benjamin Stemper. A regularity structure for rough volatility.Mathematical Finance, 30(3):782–832, 2020

    2020

  13. [13]

    Optimal stopping with signatures.The Annals of Applied Probability, 33(1):238–273, 2023

    Christian Bayer, Paul P Hager, Sebastian Riedel, and John Schoenmakers. Optimal stopping with signatures.The Annals of Applied Probability, 33(1):238–273, 2023

    2023

  14. [14]

    Primal and dual optimal stopping with signatures

    Christian Bayer, Luca Pelizzari, and John Schoenmakers. Primal and dual optimal stopping with signatures. Finance and Stochastics, 29(4):981–1014, 2025

    2025

  15. [15]

    Correlators of polynomial processes.SIAM Journal on Financial Mathe- matics, 12(4):1374–1415, 2021

    Fred Espen Benth and Silvia Lavagnini. Correlators of polynomial processes.SIAM Journal on Financial Mathe- matics, 12(4):1374–1415, 2021

    2021

  16. [16]

    The expected signature of brownian motion stopped on the boundary of a circle has finite radius of convergence.Bulletin of the London Mathematical Society, 53(1):285–299, 2021

    Horatio Boedihardjo, Joscha Diehl, Marc Mezzarobba, and Hao Ni. The expected signature of brownian motion stopped on the boundary of a circle has finite radius of convergence.Bulletin of the London Mathematical Society, 53(1):285–299, 2021

    2021

  17. [17]

    The signature of a rough path: uniqueness.Advances in Mathematics, 293:720–737, 2016

    Horatio Boedihardjo, Xi Geng, Terry Lyons, and Danyu Yang. The signature of a rough path: uniqueness.Advances in Mathematics, 293:720–737, 2016

    2016

  18. [18]

    A representation for the expected signature of brownian motion up to the first exit time of the planar unit disc

    Horatio Boedihardjo, Lin Hao He, and Lisa Wang. A representation for the expected signature of brownian motion up to the first exit time of the planar unit disc. In Dan Crisan, Ilya Chevyrev, Thomas Cass, James Foster, Christian Litterer, and Cristopher Salvi, editors,Stochastic Analysis and Applications 2025, pages 189–213. Springer, Cham, 2026

    2025

  19. [19]

    Expected signature of gaussian processes with strictly regular kernels, 2013

    Horatio Boedihardjo, Anastasia Papavasiliou, and Zhongmin Qian. Expected signature of gaussian processes with strictly regular kernels, 2013

    2013

  20. [20]

    Rough differential equations for volatil- ity.arXiv preprint arXiv:2412.21192, 2024

    Ofelia Bonesini, Emilio Ferrucci, Ioannis Gasteratos, and Antoine Jacquier. Rough differential equations for volatil- ity.arXiv preprint arXiv:2412.21192, 2024

    arXiv 2024

  21. [21]

    Adapted topologies and higher rank signatures.The Annals of Applied Probability, 33(3):2136–2175, 2023

    Patric Bonnier, Chong Liu, and Harald Oberhauser. Adapted topologies and higher rank signatures.The Annals of Applied Probability, 33(3):2136–2175, 2023

    2023

  22. [22]

    Signature cumulants, ordered partitions, and independence of stochastic processes.Bernoulli, 26(4):2727–2757, 2020

    Patric Bonnier and Harald Oberhauser. Signature cumulants, ordered partitions, and independence of stochastic processes.Bernoulli, 26(4):2727–2757, 2020

    2020

  23. [23]

    Proper scoring rules, gradients, divergences, and entropies for paths and time series.Bayesian Analysis, 20(4):1615–1646, 2025

    Patric Bonnier and Harald Oberhauser. Proper scoring rules, gradients, divergences, and entropies for paths and time series.Bayesian Analysis, 20(4):1615–1646, 2025

    2025

  24. [24]

    Double-execution strategies using path signatures.SIAM Journal on Financial Mathematics, 13(4):1379–1417, 2022

    ´Alvaro Cartea, Imanol P´ erez Arribas, and Leandro S´ anchez-Betancourt. Double-execution strategies using path signatures.SIAM Journal on Financial Mathematics, 13(4):1379–1417, 2022

    2022

  25. [25]

    On the wiener chaos expansion of the signature of a gaussian process.Probability Theory and Related Fields, 189:909–947, 2024

    Thomas Cass and Emilio Ferrucci. On the wiener chaos expansion of the signature of a gaussian process.Probability Theory and Related Fields, 189:909–947, 2024

    2024

  26. [26]

    Lyons, and Xingcheng Xu

    Thomas Cass, Terry J. Lyons, and Xingcheng Xu. Weighted signature kernels.The Annals of Applied Probability, 34(1A):585–626, 2024

    2024

  27. [27]

    Iterated integrals and exponential homomorphisms.Proceedings of the London Mathematical So- ciety, 3(1):502–512, 1954

    Kuo-Tsai Chen. Iterated integrals and exponential homomorphisms.Proceedings of the London Mathematical So- ciety, 3(1):502–512, 1954

    1954

  28. [28]

    Integration of paths, geometric invariants and a generalized baker-hausdorff formula.Annals of Mathematics, 65(1):163–178, 1957

    Kuo-Tsai Chen. Integration of paths, geometric invariants and a generalized baker-hausdorff formula.Annals of Mathematics, 65(1):163–178, 1957

    1957

  29. [29]

    Orthogonal polynomials on path-space, 2026

    Ilya Chevyrev, Emilio Ferrucci, Darrick Lee, Terry Lyons, Harald Oberhauser, and Nikolas Tapia. Orthogonal polynomials on path-space, 2026

    2026

  30. [30]

    Characteristic functions of measures on geometric rough paths

    Ilya Chevyrev and Terry Lyons. Characteristic functions of measures on geometric rough paths. 2016

    2016

  31. [31]

    Signature moments to characterize laws of stochastic processes.The Journal of Machine Learning Research, 23(1):7928–7969, 2022

    Ilya Chevyrev and Harald Oberhauser. Signature moments to characterize laws of stochastic processes.The Journal of Machine Learning Research, 23(1):7928–7969, 2022

    2022

  32. [32]

    Friz, and Harald Oberhauser

    Dan Crisan, Joscha Diehl, Peter K. Friz, and Harald Oberhauser. Robust filtering: Correlated noise and multidi- mensional observation.The Annals of Applied Probability, 23(5):2139–2160, 2013

    2013

  33. [33]

    Solving backward stochastic differential equations using the cubature method: Application to nonlinear pricing.SIAM Journal on Financial Mathematics, 3(1):534–571, 2012

    Dan Crisan and Konstantinos Manolarakis. Solving backward stochastic differential equations using the cubature method: Application to nonlinear pricing.SIAM Journal on Financial Mathematics, 3(1):534–571, 2012

    2012

  34. [34]

    Signature-Based Models: Theory and calibration.SIAM Journal on Financial Mathematics, 14(3):910–957, 2023

    Christa Cuchiero, Guido Gazzani, and Sara Svaluto-Ferro. Signature-Based Models: Theory and calibration.SIAM Journal on Financial Mathematics, 14(3):910–957, 2023

    2023

  35. [35]

    Polynomial processes and their applications to mathematical finance.Finance and Stochastics, 16(4):711–740, 2012

    Christa Cuchiero, Martin Keller-Ressel, and Josef Teichmann. Polynomial processes and their applications to mathematical finance.Finance and Stochastics, 16(4):711–740, 2012

    2012

  36. [36]

    Probability measure-valued polynomial diffusions

    Christa Cuchiero, Martin Larsson, and Sara Svaluto-Ferro. Probability measure-valued polynomial diffusions. 2019

    2019

  37. [37]

    Polynomial McKean-Vlasov SDEs.arXiv preprint arXiv:2502.19203, 2025

    Christa Cuchiero and Janka M¨ oller. Polynomial McKean-Vlasov SDEs.arXiv preprint arXiv:2502.19203, 2025

    Pith/arXiv arXiv 2025

  38. [38]

    Infinite-dimensional polynomial processes.Finance and Stochastics, 25(2):383–426, 2021

    Christa Cuchiero and Sara Svaluto-Ferro. Infinite-dimensional polynomial processes.Finance and Stochastics, 25(2):383–426, 2021

    2021

  39. [39]

    Signature sdes from an affine and polynomial perspec- tive.arXiv preprint arXiv:2302.01362, 2023

    Christa Cuchiero, Sara Svaluto-Ferro, and Josef Teichmann. Signature sdes from an affine and polynomial perspec- tive.arXiv preprint arXiv:2302.01362, 2023. EXPECTED SIGNATURES VIA PARTIAL INTEGRATION, COORDINATE CHANGE AND SYMMETRIZATION 35

    Pith/arXiv arXiv 2023

  40. [40]

    Joscha Diehl, Peter K Friz, and Wilhelm Stannat. Stochastic partial differential equations: a rough paths view on weak solutions via Feynman–Kac.Annales de la Facult´ e des sciences de Toulouse: Math´ ematiques, 26(4):911–947, 2017

    2017

  41. [41]

    Joscha Diehl, Harald Oberhauser, and Sebastian Riedel. A l´ evy area between brownian motion and rough paths with applications to robust nonlinear filtering and rough partial differential equations.Stochastic processes and their applications, 125(1):161–181, 2015

    2015

  42. [42]

    Invariants of multidimensional time series based on their iterated-integral signature.Acta Applicandae Mathematicae, 164(1):83–122, 2019

    Joscha Diehl and Jeremy Reizenstein. Invariants of multidimensional time series based on their iterated-integral signature.Acta Applicandae Mathematicae, 164(1):83–122, 2019

    2019

  43. [43]

    PhD thesis, University of Oxford, 2003

    Thomas Fawcett.Problems in Stochastic Analysis: Connections Between Rough Paths and Non-Commutative Harmonic Analysis. PhD thesis, University of Oxford, 2003

    2003

  44. [44]

    Emilio Ferrucci, Timothy Herschell, Christian Litterer, and Terry J. Lyons. High-degree cubature on wiener space through unshuffle expansions.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sci- ences, 482(2330):20250051, 01 2026

    2026

  45. [45]

    Polynomial diffusions and applications in finance.Finance and Stochastics, 20(4):931–972, 2016

    Damir Filipovi´ c and Martin Larsson. Polynomial diffusions and applications in finance.Finance and Stochastics, 20(4):931–972, 2016

    2016

  46. [46]

    Polynomial jump-diffusion models.Stochastic Systems, 10(1):71–97, 2020

    Damir Filipovi´ c and Martin Larsson. Polynomial jump-diffusion models.Stochastic Systems, 10(1):71–97, 2020

    2020

  47. [47]

    Rough semimartingales andp-variation estimates for martingale transforms

    Peter Friz and Pavel Zorin-Kranich. Rough semimartingales andp-variation estimates for martingale transforms. arXiv preprint arXiv:2008.08897, 2020

    Pith/arXiv arXiv 2008

  48. [48]

    Friz and Paul P

    Peter K. Friz and Paul P. Hager. Expected signature kernels for l´ evy rough paths, 2025

    2025

  49. [49]

    Friz, Paul P

    Peter K. Friz, Paul P. Hager, and Nikolas Tapia. Unified signature cumulants and generalized Magnus expansions. Forum of Mathematics, Sigma, 10:e42, 2022

    2022

  50. [50]

    Friz, Paul P

    Peter K. Friz, Paul P. Hager, and Nikolas Tapia. On expected signatures and signature cumulants in semimartingale models. In Christian Bayer, Gon¸ calo dos Reis, Blanka Horvath, and Harald Oberhauser, editors,Signature Methods in Finance: An Introduction with Computational Applications, pages 381–424. Springer, Cham, 2026

    2026

  51. [51]

    Springer, 2020

    Peter K Friz and Martin Hairer.A course on rough paths. Springer, 2020

    2020

  52. [52]

    Rough stochastic differential equations.arXiv preprint arXiv:2106.10340, 2021

    Peter K Friz, Antoine Hocquet, and Khoa Lˆ e. Rough stochastic differential equations.arXiv preprint arXiv:2106.10340, 2021

    arXiv 2021

  53. [53]

    Friz and Atul Shekhar

    Peter K. Friz and Atul Shekhar. General rough integration, L´ evy rough paths and a L´ evy–kintchine-type formula. The Annals of Probability, 45(4):2707–2765, 2017

    2017

  54. [54]

    Cambridge University Press, 2010

    Peter K Friz and Nicolas B Victoir.Multidimensional stochastic processes as rough paths: theory and applications, volume 120. Cambridge University Press, 2010

    2010

  55. [55]

    A partial rough path space for rough volatility.Electronic Journal of Prob- ability, 29:1–28, 2024

    Masaaki Fukasawa and Ryoji Takano. A partial rough path space for rough volatility.Electronic Journal of Prob- ability, 29:1–28, 2024

    2024

  56. [56]

    Signature trading: A path-dependent extension of the mean- variance framework with exogenous signals, 2023

    Owen Futter, Blanka Horvath, and Magnus Wiese. Signature trading: A path-dependent extension of the mean- variance framework with exogenous signals, 2023

    2023

  57. [57]

    Springer, New York, 2 edition, 2011

    Jean Gallier.Geometric Methods and Applications: For Computer Science and Engineering, volume 38 ofTexts in Applied Mathematics. Springer, New York, 2 edition, 2011

    2011

  58. [58]

    Gaussian rough paths lifts via complementary Young regularity.Electronic Commu- nications in Probability, 29:1–13, 2024

    Paul Gassiat and Tom Klose. Gaussian rough paths lifts via complementary Young regularity.Electronic Commu- nications in Probability, 29:1–13, 2024

    2024

  59. [59]

    Expected signature on a riemannian manifold and its geometric implications,

    Xi Geng, Hao Ni, and Chaorui Wang. Expected signature on a riemannian manifold and its geometric implications,

  60. [60]

    Differential equations driven by Π-rough paths.Proceedings of the Edinburgh Mathematical Society, 59(3):741–758, 2016

    Lajos Gergely Gyurk´ o. Differential equations driven by Π-rough paths.Proceedings of the Edinburgh Mathematical Society, 59(3):741–758, 2016

    2016

  61. [61]

    Lajos Gergely Gyurk´ o and Terry J. Lyons. Efficient and practical implementations of cubature on wiener space. In Dan Crisan, editor,Stochastic Analysis 2010, pages 73–111. Springer, Berlin, Heidelberg, 2011

    2010

  62. [62]

    Hager and Luca Pelizzari

    Paul P. Hager and Luca Pelizzari. A technical note on signature computations with mixed truncation, coordinate changes, and symmetrization.https://github.com/hagerpa/tensordev/academia/bidegree, 2026. Working note

    2026

  63. [63]

    Uniqueness for the signature of a path of bounded variation and the reduced path group.Ann

    Ben Hambly and Terry Lyons. Uniqueness for the signature of a path of bounded variation and the reduced path group.Ann. of Math. (2), 171(1):109–167, 2010

    2010

  64. [64]

    Monte carlo construction of cubature on wiener space.Japan Journal of Industrial and Applied Mathematics, 39(2):543–571, 2022

    Satoshi Hayakawa and Ken’ichiro Tanaka. Monte carlo construction of cubature on wiener space.Japan Journal of Industrial and Applied Mathematics, 39(2):543–571, 2022

    2022

  65. [65]

    Optimal execution with rough path signatures.SIAM Journal on Financial Mathematics, 11(2):470–493, 2020

    Jasdeep Kalsi, Terry Lyons, and Imanol Perez Arribas. Optimal execution with rough path signatures.SIAM Journal on Financial Mathematics, 11(2):470–493, 2020

    2020

  66. [66]

    Kloeden and Eckhard Platen.Numerical Solution of Stochastic Differential Equations, volume 23 of Stochastic Modelling and Applied Probability

    Peter E. Kloeden and Eckhard Platen.Numerical Solution of Stochastic Differential Equations, volume 23 of Stochastic Modelling and Applied Probability. Springer, Berlin, Heidelberg, 1992

    1992

  67. [67]

    On (p, q)-rough paths.Journal of Differential Equations, 225(1):103–133, 2006

    Antoine Lejay and Nicolas Victoir. On (p, q)-rough paths.Journal of Differential Equations, 225(1):103–133, 2006

    2006

  68. [68]

    Bonilla, and Terry Lyons

    Maud Lemercier, Cristopher Salvi, Theodoros Damoulas, Edwin V. Bonilla, and Terry Lyons. Distribution re- gression for sequential data. InProceedings of the 24th International Conference on Artificial Intelligence and Statistics, volume 130 ofProceedings of Machine Learning Research, pages 3754–3762. PMLR, 2021

    2021

  69. [69]

    Learning from the past, predicting the statistics for the future, learning an evolving system, 2013

    Daniel Levin, Terry Lyons, and Hao Ni. Learning from the past, predicting the statistics for the future, learning an evolving system, 2013. 36 PAUL P. HAGER AND LUCA PELIZZARI

    2013

  70. [70]

    Expected signature of stopped brownian motion ond-dimensionalC 2,α-domains has finite radius of convergence everywhere: 2≤d≤8.Journal of Functional Analysis, 282(12):109447, 2022

    Siran Li and Hao Ni. Expected signature of stopped brownian motion ond-dimensionalC 2,α-domains has finite radius of convergence everywhere: 2≤d≤8.Journal of Functional Analysis, 282(12):109447, 2022

    2022

  71. [71]

    Sig-Wasserstein GANs for conditional time series generation.Mathematical Finance, 34(2):622–670, 2024

    Shujian Liao, Hao Ni, Marc Sabate-Vidales, Lukasz Szpruch, Magnus Wiese, and Baoren Xiao. Sig-Wasserstein GANs for conditional time series generation.Mathematical Finance, 34(2):622–670, 2024

    2024

  72. [72]

    High order recombination and an application to cubature on Wiener space

    Christian Litterer and Terry Lyons. High order recombination and an application to cubature on Wiener space. The Annals of Applied Probability, 22(4):1301–1327, 2012

    2012

  73. [73]

    Pakkanen, and Almut E

    Lorenzo Lucchese, Mikko S. Pakkanen, and Almut E. D. Veraart. Learning with expected signatures: Theory and applications. InProceedings of the 42nd International Conference on Machine Learning, volume 267 ofProceedings of Machine Learning Research, pages 40995–41055. PMLR, 2025

    2025

  74. [74]

    Numerical method for model-free pricing of exotic derivatives in discrete time using rough path signatures.Applied Mathematical Finance, 26(6):583–597, 2019

    Terry Lyons, Sina Nejad, and Imanol Perez Arribas. Numerical method for model-free pricing of exotic derivatives in discrete time using rough path signatures.Applied Mathematical Finance, 26(6):583–597, 2019

    2019

  75. [75]

    Expected signature of brownian motion up to the first exit time from a bounded domain

    Terry Lyons and Hao Ni. Expected signature of brownian motion up to the first exit time from a bounded domain. The Annals of Probability, 43(5):2729–2762, 2015

    2015

  76. [76]

    Cubature on wiener space.Proceedings of the Royal Society of London

    Terry Lyons and Nicolas Victoir. Cubature on wiener space.Proceedings of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences, 460(2041):169–198, 2004

    2041

  77. [77]

    Differential equations driven by rough signals.Revista Matem´ atica Iberoamericana, 14(2):215–310, 1998

    Terry J Lyons. Differential equations driven by rough signals.Revista Matem´ atica Iberoamericana, 14(2):215–310, 1998

    1998

  78. [78]

    Signature-based optimal execution for statistical arbitrage with path-dependent trading signals.arXiv preprint arXiv:2606.31387, 2026

    Gianmarco Morbelli, Sven Karbach, and Mike Derksen. Signature-based optimal execution for statistical arbitrage with path-dependent trading signals.arXiv preprint arXiv:2606.31387, 2026

    Pith/arXiv arXiv 2026

  79. [79]

    PhD thesis, Oxford University, UK, 2012

    Hao Ni.The expected signature of a stochastic process. PhD thesis, Oxford University, UK, 2012

    2012

  80. [80]

    Concentration and exact convergence rates for expected brownian signatures.Electronic Communications in Probability, 20(8):1–11, 2015

    Hao Ni and Weijun Xu. Concentration and exact convergence rates for expected brownian signatures.Electronic Communications in Probability, 20(8):1–11, 2015

    2015

Showing first 80 references.

This paper was first reviewed by deepseek-v4-flash on August 3, 2026.