Pith. sign in

REVIEW 2 major objections 4 minor 97 references

Dynamic Universal Approximation via Signature Controlled Differential Equations

T0 review · 2 major / 4 minor · reviewed 2026-08-02 · deepseek-v4-flash

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

Pith's one-line read Every well-posed path-dependent controlled differential equation can be approximated arbitrarily well by finite-dimensional signature-controlled equations whose vector fields have the elementary form of an activation applied to a linear fun

desk verdict Dynamic universality for Sig-CDEs is real and the main proof holds up; the result is new, with a few minor blemishes and one legitimate external dependency. read the letter →

arxiv 2607.13886 v1 pith:6KYQIBCP submitted 2026-07-15 math.CA math.PR

Tomás Carrondo , Christa Cuchiero , Paul P. Hager , Fabian N. Harang This is my paper
classification math.CAmath.PR MSC 34K0560L1034G2041A65
keywords path-dependentcontrolleddifferentialequationspathsignaturesdynamicuniversalapproximationHölderspacessignatureCDEsweightedinfinite-dimensionalstate-spaceliftsnilpotentLiegroups
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that any well-posed path-dependent controlled differential equation — any system whose current rate of change depends on the entire past trajectory — can be replaced, to arbitrary accuracy, by a finite-dimensional equation in which the memory is encoded by the truncated path signature and the vector field is a scalar activation of a linear functional of that signature. If true, this gives a canonical, learnable, finite-dimensional surrogate for infinite-dimensional hereditary dynamics: the same parameterization works uniformly over bounded sets of controls and initial histories, and even globally over controls in a weighted sense. The proof connects a new well-posedness theory for path-dependent CDEs on Hölder spaces with the classical universality of signatures, and transfers approximation of vector fields to approximation of solutions via a stability estimate.

What carries the argument

The signature map S(y)_{0,t} — the graded collection of iterated integrals of a path — is the memory encoding device: truncated to level N it solves a linear CDE dS = S ⊗ dy and places the history in the finite-dimensional step-N nilpotent Lie group. The argument's engine is the weighted universality of signature linear functionals: on Hölder path spaces equipped with an exponential weight, every sufficiently regular path functional is approximated by ℓ ↦ ⟨ℓ, S⟩. This static approximation is converted into dynamic approximation by (i) passing through a strictly monotone logarithmic-growth activation σ to enforce the linear-growth condition needed for well-posedness, and (ii) the stability es

What would settle it

Take a concrete non-anticipative f satisfying (26)-(27), such as a delayed feedback term f(y|_{[0,t]})=sin(y_{t-r}), and search numerically for the ℓ̂_n sequence: if the best uniform signature-linear approximation under the exponential weight forces ∥ℓ̂_n∥ to grow faster than the logarithmic bound in (47), the uniform a priori bound on y^n breaks and Theorem 4.7 fails. A direct contradiction would be a pair (f, x, w) satisfying the hypotheses for which the solutions of (41) do not converge in β-Hölder norm.

Watch

Extended reading notes

Core claim

The paper's central claim is Theorem 4.7: for any continuous non-anticipative vector field f that is Lipschitz on compact sets and has linear growth on Hölder path spaces, the solution y of the path-dependent CDE dy_t = f(y|_{[0,t]}) dx_t is the β-Hölder limit of solutions y^n of signature CDEs whose readouts are restricted to c_0 + c_1 σ(⟨ℓ^n, S(ŷ|_{[0,s]})⟩). The convergence is uniform over bounded sets of Lipschitz controls and α-Hölder initial histories. The paper further shows a global version on weighted spaces (Theorem 4.11), well-posedness of the infinite-dimensional 'lifted' equation in projective limits of tensor spaces (Theorem 4.16), and intrinsic global well-posedness on step-N

Load-bearing premise

The whole construction collapses if the imported static universality theorem fails for the exponential weight ψ̂: the approximating coefficients ℓ̂_n are produced by that theorem, and nothing else in the paper supplies them.

Editorial extensions

If this is right

  • Any well-posed path-dependent CDE admits finite-dimensional approximations that are differentiable in their parameters, so gradient-based learning of the coefficients is possible.
  • The approximating latent state is a truncated signature, hence a continuous feature map of the observed trajectory in Hölder norm; this provides stability with respect to noisy, oscillatory, or irregularly sampled inputs.
  • Uniform approximation over bounded sets of controls and initial histories means one set of coefficients works for a whole family of input signals, with a global version holding over all controls in a weighted sense.
  • The lifted infinite-dimensional picture provides a canonical state-space form for path-dependent dynamics, with explicit tensor-level conditions that parallel classical delay-equation semigroup theory.
  • On step-N nilpotent Lie groups, intrinsic global existence holds under homogeneous linear growth, but intrinsic Lipschitz conditions do not guarantee uniqueness, delimiting where intrinsic formulations can be used safely.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A direct corollary left implicit: the theorem supplies a theoretical justification for replacing recurrent or path-dependent neural architectures by signature-CDEs with a fixed signature feature map; one could test this on benchmark delay-differential systems.
  • The projective-limit tensor construction suggests a topological reading of 'how many signature levels are needed': membership in T^{(p)} encodes levelwise decay, and one could empirically measure the convergence rate in the truncation level N for specific functionals and compare it with the λ_k/k^p condition.
  • The authors note that a stochastic analogue is the subject of accompanying work; if the dynamic universality transfers to stochastic drivers, the same parameterization would yield universal approximations for path-dependent SDEs, but the pathwise Hölder estimates would need to be replaced by probabilistic ones.
  • The proof's reliance on a weighted universal approximation theorem means the choice of weight function directly controls how large a control ball can be covered; exploring subexponential weights could sharpen the global control-level result.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper develops a well-posedness and approximation theory for path-dependent controlled differential equations (CDEs) on spaces of stopped α-Hölder paths. The main result, Theorem 4.7, asserts that for any continuous non-anticipative vector field f satisfying local Lipschitz and linear growth conditions, the solution of d y_t = f(y\|_{[0,t]}) dx_t can be approximated in β-Hölder norm, uniformly over bounded sets of Lipschitz controls and initial histories, by solutions of finite-dimensional signature CDEs whose readouts have the restricted form c0 + c1 σ(⟨ℓ, S(ŷ)⟩). The proof combines the static weighted universality of signature linear functionals (Theorem 2.7, imported from [35]) with an explicit stability estimate for path-dependent CDEs (Theorem 3.10). A global control-level variant is stated as Theorem 4.11. The paper also constructs projective-limit tensor spaces T^{(p)}, proves well-posedness for infinite-dimensional lifted Sig-CDEs (Theorem 4.16), and studies truncated Sig-CDEs on step-N groups under intrinsic regularity (Theorem 4.30), including an equivalence between path-level and Euclidean Lipschitz regularity (Proposition 4.34).

Significance. If Theorem 4.7 is accepted, this is a substantial contribution: it gives a genuinely dynamic universality statement for signature-based models under weak assumptions, with explicit rates via stability estimates and uniform a priori bounds that avoid the usual 'compact set depends on approximants' circularity. The explicit stability estimate in Theorem 3.10, the uniform a priori bound in (47), and the newly introduced T^{(p)} spaces are useful technical ingredients. The paper is largely self-contained after importing Theorem 2.7 from [35]; I do not regard that import as circular, since the static statement concerns functionals and does not contain the dynamic result. However, the global variant (Theorem 4.11) and the intrinsic group theorem (4.30) have proof gaps described below, so the full claims of the paper are not yet established.

major comments (2)
  1. [§4.2, proof of Theorem 4.11 (around Eq. (50))] The proof fixes the weighted control space ψ_c before the approximating ℓ is chosen, and the a priori bound for y^Sig is derived 'for any functional ℓ' with a constant that, by the authors' own parenthetical, 'depends inclusively on ℓ'. The constants k_6,k_7 in the subsequent R_Sig expression are therefore not shown to be ℓ-independent. If they are ℓ-dependent, then R_Sig(∥x∥), which enters (50) and the definition of η_c, is not known until after ℓ is selected, and the claimed global weighted estimate can fail for the ℓ actually produced. Since Theorem 4.11 is the advertised global variant, this is load-bearing. A repair would be to use the B_{ψ̂}-bound on ℓ∘S from (44) to obtain an ℓ-independent R_Sig for all ℓ satisfying the approximation inequality, in the same way (47) is used in the proof of Theorem 4.7, and only then fix ψ_c.
  2. [§4.5, proof of Theorem 4.30 (after Eq. (77))] The proof reduces (70) to the log-coordinate equation (72) and estimates |Ω_t| ≲ |Ω_0| + ∫(1+∥Ω_u∥^N)|dx|. This is a superlinear Gronwall inequality; Grönwall's lemma does not yield global existence for y' ≤ C(1+y^N). The theorem may be salvageable by exploiting the triangular structure of H to argue level-by-level with the homogeneous growth of F, or by estimating the homogeneous gauge directly on the group, but as written the global-existence claim is not proved. The second (level-wise triangular) part of the theorem uses Bihari–LaSalle and is not affected by this objection.
minor comments (4)
  1. [§4.2, proof of Theorem 4.7, after Eq. (45)] The sentence 'which, by Lemma A.3, is equivalent to ∥f^{ij} − ...∥_{Bψ} < ε_n' is not literally correct: the Tietze extension \hat f^{ij} is not the lift of f^{ij}, so Lemma A.3 does not transfer the weighted inequality from the extension to the original functional. The subsequent use of the estimate on the compact set K, where the extension agrees with f^{ij}, is valid, so this is a harmless but misleading remark.
  2. [§4.2, Theorems 4.7 and 4.11] The statements allow the initial time T0 to vary, but the stability theorem (Theorem 3.10) compares solutions with the same T0. Either T0 should be fixed throughout, or the uniformity over variable T0 should be justified explicitly.
  3. [§4.2, proof of Theorem 4.7] The choice of c0,c1 so that σ^{-1}(c1^{-1}\hat f^{ij} − c0 c1^{-1}) is defined is asserted without detail. Since a Lipschitz, strictly monotone activation satisfying (39) may have bounded range, the proof should state that c1 is chosen large enough that the bounded range of \hat f^{ij} is mapped into the interior of σ(R) uniformly in i,j.
  4. [§4.2, Theorem 4.11, Eq. (50)] The constant M is defined as (inf_x ψ_c(x))^{-1}; this requires inf_x ψ_c(x) > 0. This holds for the explicit η_c constructed at the end of the proof, but it would be clearer to state it before using M in (50).

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the dynamic universality proof rests on an independently stated weighted-signature universality theorem; the overlapping author citation is minor and not a circular reduction.

full rationale

The claimed derivation chain is not circular. Theorem 3.1 is proved from scratch by Schauder fixed-point, compactness and Grönwall-type arguments, with Section 4 reducing Sig-CDE well-posedness to it. Theorem 4.7 does not fit the coefficients ℓ_n to the target solution; it constructs them by applying the external static universality result Theorem 2.7 to a bounded extension of a σ^{-1}-transformed f^{ij}, and then transfers the vector-field approximation to solution paths through the stability estimate Theorem 3.10. The only self-cited input is Theorem 2.7, whose proof the paper refers to as identical to [35, Theorem 5.4]. Although [35] shares an author and is load-bearing in the sense that the construction of ℓ_n would fail if the weighted universality were false, it is a published, parameter-free static statement about approximating functionals on weighted Hölder path spaces by signature linear functionals; it does not assume the dynamic solution approximation theorem. The exponential-weight admissibility is checked through compact Hölder embeddings (Examples 2.2 and 2.4), so this is an external dependency, not a self-referential reduction. The proof also explicitly avoids the circular dependence between the compact set and the approximants: it obtains a global weighted approximation first, then uses (47) to get an n-uniform a priori bound and only then fixes the common compact set K. Theorem 4.11's phrase 'with no circularity' is likewise justified: the constants k1, k2, k6, k7 are independent of Bx and the a posteriori choice of k0 is a routine quantifier-ordering step for a fixed δ. The only proof-level caveats—Theorem 2.7 is imported as a black box, and the Lemma A.3 norm-transfer sentence is imprecise—do not turn the dynamic result into its own input. Hence no genuine circular step; the minor self-citation supports a low non-circularity score of 2.

Assumptions & free parameters 3 free parameters · 6 assumptions · 1 invented entities

The dynamic universal approximation theorem depends on the imported static universality of signatures (Theorem 2.7 from prior work with overlapping authors) and on standard rough-path/FDE machinery. The new T(p) spaces are auxiliary but load-bearing for the infinite-dimensional lift; no empirical free parameters are fitted.

free parameters (3)
  • p (exponent in T^{(p)} state spaces)
    Any p>1 works; the construction requires p>1 so the tensor product is continuous and compact embeddings hold (Lemmas 2.14, 2.15). Not fitted to data.
  • weight sequence (λ_k) in T^{(p)}
    Positive sequence with λ_k^{1/β}/k^p → 0 (Prop 2.10) used to define the projective limit; existence suffices, any such sequence works.
  • weight parameters ζ>0, ξ≥1 in ψ̂
    Define the exponential weight in the weighted universality theorem (Thm 2.7); arbitrary, only existence is needed.
assumptions (6)
  • domain assumption Global weighted universality of signature linear functionals (Theorem 2.7, imported from Cuchiero–Schmocker–Teichmann [35])
    Used as the black-box approximation engine in the proof of Theorem 4.7 (Eq. (44)).
  • standard math Lyons' extension and factorial decay of signatures (Thm 3.1.3 in [71])
    Provides the local Lipschitz continuity of S into T(p) (Prop 2.10) and the growth estimate (9).
  • standard math Schauder fixed point theorem (Theorem 3.3)
    Key tool for local existence of path-dependent CDEs (Prop 3.4).
  • standard math Young/Riemann-Stieltjes integration for α-Hölder paths with α>1/2
    Signatures and CDEs are defined via Young integrals; see Section 2.2 and Remark 2.1.
  • standard math Compact embedding C^{α-Hölder} ↪ C^{β-Hölder} for β<α
    Used to make weighted spaces admissible and to apply Arzelà-Ascoli; see Examples 2.2, 2.4 and Theorem 4.16.
  • domain assumption Time-augmentation separates stopped paths (Remark 2.6)
    Needed so that signature linear functionals are point-separating on the path space used in the universality theorem.
invented entities (1)
  • Projective limit tensor spaces T^{(p)}(R^m) (and T(R^m))
    purpose: Banach spaces (resp. Fréchet-type limits) serving as codomain of the signature map and state space for lifted Sig-CDEs; their strong topologies let F be merely locally Lipschitz while preserving signature regularity
    Introduced in Section 2.4; Remark 2.13 notes they appear not to have been explicitly identified in the literature. No external falsifiable predictions; they are definitions used inside the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Dynamic Universal Approximation via Signature Controlled Differential Equations." pith.science (2026). https://pith.science/paper/6KYQIBCP

@misc{pith2026260713886,
  author       = {Pith},
  title        = {Pith review of: Dynamic Universal Approximation via Signature Controlled Differential Equations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6KYQIBCP}},
  note         = {Machine review of arXiv:2607.13886}
}
abstract

We study signature controlled differential equations (Sig-CDEs), that is, path-dependent controlled differential equations (CDEs) whose vector fields factor through the signature map. Working on spaces of stopped H\"older paths, we develop an existence, uniqueness, and stability theory for general path-dependent CDEs, and translate these pathwise well-posedness criteria into conditions on the corresponding signature functionals. We then prove dynamic universality: simply parametrized Sig-CDEs approximate the solution path of any well-posed path-dependent CDE arbitrarily well, uniformly over bounded sets of controls and initial histories, with global variants obtained using weighted spaces. Within this framework, entire maps of group-like elements provide a specific class of Sig-CDEs. Using a new class of limiting tensor spaces, we recast Sig-CDEs as infinite-dimensional classical CDEs and prove their well-posedness via a gauge-type scaling argument, thereby establishing a principled way to lift generic path-dependent dynamics. Lastly, we study truncated Sig-CDEs as finite-dimensional differential equations on step$-N$ Lie groups under intrinsic conditions, that is, with well-posedness formulated in terms of the underlying group metric.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

97 extracted references · 6 linked inside Pith

  1. [35]

    Global universal approximation of functional input maps on weighted spaces.Constructive Approximation, pages 1–76, 2026

    Christa Cuchiero, Philipp Schmocker, and Josef Teichmann. Global universal approximation of functional input maps on weighted spaces.Constructive Approximation, pages 1–76, 2026

    2026

  2. [1]

    Martingale property and moment explosions in signature volatility models.arXiv Preprint: 2503.17103, 2025

    Eduardo Abi Jaber, Paul Gassiat, and Dimitri Sotnikov. Martingale property and moment explosions in signature volatility models.arXiv Preprint: 2503.17103, 2025

    arXiv 2025

  3. [2]

    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. [3]

    Exponentially fading memory signature.arXiv Preprint: 2507.03700, 2025

    Eduardo Abi Jaber and Dimitri Sotnikov. Exponentially fading memory signature.arXiv Preprint: 2507.03700, 2025

    arXiv 2025

  5. [4]

    Cambridge University Press, 2019

    Andrei Agrachev, Davide Barilari, and Ugo Boscain.A Comprehensive Introduction to Sub-Riemannian Geometry, volume 181 ofCambridge Studies in Advanced Mathematics. Cambridge University Press, 2019

    2019

  6. [5]

    Agrachev and Yuri L

    Andrei A. Agrachev and Yuri L. Sachkov.Control Theory from the Geometric Viewpoint., volume 87 ofEncycl. Math. Sci.Berlin: Springer, 2004

    2004

  7. [6]

    Rough differential equations containing path-dependent bounded variation terms.Journal of Theoretical Probability, 37(3):2130–2183, 2024

    Shigeki Aida. Rough differential equations containing path-dependent bounded variation terms.Journal of Theoretical Probability, 37(3):2130–2183, 2024

    2024

  8. [7]

    Aliprantis and Kim C

    Charalambos D. Aliprantis and Kim C. Border.Infinite Dimensional Analysis: a Hitchhiker’s Guide. Springer, 3rd edition, 2006

    2006

Show all 97 references
  1. [8]

    Rough differential equations with path-dependent coefficients.Annales Henri Lebesgue, 6:1–29, 2023

    Anna Ananova. Rough differential equations with path-dependent coefficients.Annales Henri Lebesgue, 6:1–29, 2023

    2023

  2. [9]

    Path dependent equations driven by H¨ older processes.Stochastic Analysis and Applications, 37(3):480–498, 2019

    Rafael Andretto Castrequini and Francesco Russo. Path dependent equations driven by H¨ older processes.Stochastic Analysis and Applications, 37(3):480–498, 2019

    2019

  3. [10]

    Sig-SDEs model for quantitative finance

    Imanol Perez Arribas, Cristopher Salvi, and Lukasz Szpruch. Sig-SDEs model for quantitative finance. InProceedings of the First ACM International Conference on AI in Finance, pages 1–8, 2020. 10Both pathsy 1 andy 2 can be chosen to start at 0. 54

    2020

  4. [11]

    Asmar and Loukas Grafakos.Complex Analysis with Applications

    Nakhl´ e H. Asmar and Loukas Grafakos.Complex Analysis with Applications. Springer, 2018

    2018

  5. [12]

    Baum and Ted Petrie

    Leonard E. Baum and Ted Petrie. Statistical inference for probabilistic functions of finite state markov chains.The Annals of Mathematical Statistics, 37(6):1554–1563, 1966

    1966

  6. [13]

    Hager, Sebastian Riedel, and John Schoenmakers

    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

  7. [14]

    An introduction to tensors for path signatures.arXiv preprint arXiv:2502.15703, 2025

    Jack Beda, Gon¸ calo dos Reis, and Nikolas Tapia. An introduction to tensors for path signatures.arXiv preprint arXiv:2502.15703, 2025

    2025 arXiv

  8. [15]

    Friz, Sylvie Paycha, and Rosa Preiß

    Carlo Bellingeri, Peter K. Friz, Sylvie Paycha, and Rosa Preiß. Smooth rough paths, their geometry and algebraic renormalization.Vietnam J. Math., 50(3):719–761, 2022

    2022

  9. [16]

    Delfour, and Sanjoy K

    Alain Bensoussan, Giuseppe Da Prato, Michel C. Delfour, and Sanjoy K. Mitter.Represen- tation and Control of Infinite Dimensional Systems. Systems & Control: Foundations & Applications. Birkh¨ auser Boston, Boston, MA, 2 edition, 2007

    2007

  10. [17]

    A generalization of a lemma of Bellman and its application to uniqueness problems of differential equations.Acta Mathematica Hungarica, 7(1):81–94, 1956

    Imre Bihari. A generalization of a lemma of Bellman and its application to uniqueness problems of differential equations.Acta Mathematica Hungarica, 7(1):81–94, 1956

    1956

  11. [18]

    SL2(R)-developments and signature asymptotics for planar paths with bounded variation.Rev

    Horatio Boedihardjo and Xi Geng. SL2(R)-developments and signature asymptotics for planar paths with bounded variation.Rev. Mat. Iberoam., 39(5):1973–2006, 2023

    1973

  12. [19]

    Horatio Boedihardjo, Xi Geng, and Nikolaos P. Souris. Path developments and tail asymp- totics of signature for pure rough paths.Adv. Math., 364:48, 2020. Id/No 107043

    2020

  13. [20]

    Zur Theorie der elastischen Nachwirkung.Sitzungsberichte der Kaiser- lichen Akademie der Wissenschaften, 70:275–306, 1874

    Ludwig Boltzmann. Zur Theorie der elastischen Nachwirkung.Sitzungsberichte der Kaiser- lichen Akademie der Wissenschaften, 70:275–306, 1874. Foundational paper on viscoelasticity, introducing the concept of physical memory in materials

  14. [21]

    Brockett

    Roger W. Brockett. Volterra series and geometric control theory.Automatica, 12(2):167–176, 1976

    1976

  15. [22]

    Thomas Cass and William F. Turner. Topologies on unparameterised path space.Journal of Functional Analysis, 286(4):110261, 2024

    2024

  16. [23]

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

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

    1954

  17. [24]

    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

  18. [25]

    Iterated path integrals.Bulletin of the American Mathematical Society, 83(5):831–879, 1977

    Kuo-Tsai Chen. Iterated path integrals.Bulletin of the American Mathematical Society, 83(5):831–879, 1977

    1977

  19. [26]

    A primer on the signature method in machine learning.arXiv preprint arXiv:1603.03788, 2016

    Ilya Chevyrev and Andrey Kormilitzin. A primer on the signature method in machine learning.arXiv preprint arXiv:1603.03788, 2016. 55

    2016

  20. [27]

    Characteristic functions of measures on geometric rough paths.Ann

    Ilya Chevyrev and Terry Lyons. Characteristic functions of measures on geometric rough paths.Ann. Probab., 44(6):4049–4082, 2016

    2016

  21. [28]

    Functional Itˆ o calculus and stochastic integral representation of martingales.Ann

    Rama Cont and David-Antoine Fourni´ e. Functional Itˆ o calculus and stochastic integral representation of martingales.Ann. Probab., 41(1):109–133, 2013

    2013

  22. [29]

    On the support of solutions to stochastic differential equations with path-dependent coefficients.Stochastic Processes Appl., 130(5):2639–2674, 2020

    Rama Cont and Alexander Kalinin. On the support of solutions to stochastic differential equations with path-dependent coefficients.Stochastic Processes Appl., 130(5):2639–2674, 2020

    2020

  23. [30]

    Sur certaines ´ equations fonctionnelles de Volterra.Funkc

    Constantin Corduneanu. Sur certaines ´ equations fonctionnelles de Volterra.Funkc. Ekvacioj, 9:119–127, 1966

    1966

  24. [31]

    Joint calibration to SPX and VIX options with signature-based models.Mathematical Finance, 35(1):161–213, 2025

    Christa Cuchiero, Guido Gazzani, Janka M¨ oller, and Sara Svaluto-Ferro. Joint calibration to SPX and VIX options with signature-based models.Mathematical Finance, 35(1):161–213, 2025

    2025

  25. [32]

    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

  26. [33]

    Signature methods in stochastic portfolio theory.SIAM Journal on Financial Mathematics, 16(4):1239–1303, 2025

    Christa Cuchiero and Janka M¨ oller. Signature methods in stochastic portfolio theory.SIAM Journal on Financial Mathematics, 16(4):1239–1303, 2025

    2025

  27. [34]

    Universal approximation theorems for continuous functions of c` adl` ag paths and L´ evy-type signature models.Finance and Stochastics, 29(2):289–342, 2025

    Christa Cuchiero, Francesca Primavera, and Sara Svaluto-Ferro. Universal approximation theorems for continuous functions of c` adl` ag paths and L´ evy-type signature models.Finance and Stochastics, 29(2):289–342, 2025

    2025

  28. [36]

    Signature SDEs from an affine and polynomial perspective.forthcoming in the Electronic Journal of Probability, 2026

    Christa Cuchiero, Sara Svaluto-Ferro, and Josef Teichmann. Signature SDEs from an affine and polynomial perspective.forthcoming in the Electronic Journal of Probability, 2026

    2026

  29. [37]

    A semigroup point of view on splitting schemes for stochastic (partial) differential equations.arXiv preprint arXiv:1011.2651, 2010

    Philipp D¨ orsek and Josef Teichmann. A semigroup point of view on splitting schemes for stochastic (partial) differential equations.arXiv preprint arXiv:1011.2651, 2010

    2010 arXiv

  30. [38]

    R. M. Dudley and R. Norvaiˇ sa.An Introduction to p-variation and Young Integrals. With Emphasis on Sample Functions of Stochastic Processes. Aarhus: University of Aarhus, Department of Mathematical Sciences, MaPhySto, 1998

    1998

  31. [39]

    Duffaut Espinosa, Kurusch Ebrahimi-Fard, W

    Luis A. Duffaut Espinosa, Kurusch Ebrahimi-Fard, W. Steven Gray, and G. S. Venkatesh. What is... a Chen–Fliess series?Notices of the American Mathematical Society, 73(2):144– 149, 2026

    2026

  32. [40]

    Functional Itˆ o calculus.Quantitative Finance, 19(5):721–729, 2019

    Bruno Dupire. Functional Itˆ o calculus.Quantitative Finance, 19(5):721–729, 2019

    2019

  33. [41]

    Jeffrey L. Elman. Finding structure in time.Cognitive Science, 14(2):179–211, 1990. 56

    1990

  34. [42]

    Fonctionnelles causales non lin´ eaires et ind´ etermin´ ees non commutatives

    Michel Fliess. Fonctionnelles causales non lin´ eaires et ind´ etermin´ ees non commutatives. Bulletin de la Soci´ et´ e Math´ ematique de France, 109:3–40, 1981

    1981

  35. [43]

    G. B. Folland and Elias M. Stein.Hardy Spaces on Homogeneous Groups, volume 28 of Math. Notes (Princeton). Princeton University Press, Princeton, NJ, 1982

    1982

  36. [44]

    Columbia University, 2010

    David-Antoine Fournie.Functional Itˆ o Calculus and Applications. Columbia University, 2010

    2010

  37. [45]

    Friz and Paul P

    Peter K. Friz and Paul P. Hager. Expected signature kernels for L´ evy rough paths.arXiv Preprint: 2509.07893, 2025

    2025 arXiv

  38. [46]

    Friz and Martin Hairer.A Course on Rough Paths

    Peter K. Friz and Martin Hairer.A Course on Rough Paths. With an Introduction to Regularity Structures. Springer, 2nd edition, 2020

    2020

  39. [47]

    Friz and Nicolas B

    Peter K. Friz and Nicolas B. Victoir.Multidimensional Stochastic Processes as Rough Paths. Theory and Applications, volume 120 ofCambridge Studies in Advanced Mathematics. Cambridge University Press, Cambridge, 2010

    2010

  40. [48]

    Kilbas, Francesco Mainardi, and Sergei V

    Rudolf Gorenflo, Anatoly A. Kilbas, Francesco Mainardi, and Sergei V. Rogosin.Mittag- Leffler Functions, Related topics and Applications. Springer Monographs in Mathematics. Berlin: Springer, 2nd edition, 2020

    2020

  41. [49]

    Hale and M.A

    Jack K. Hale and M.A. Cruz. Existence, uniqueness and continuous dependence for hereditary systems.Annali di matematica pura ed applicata, 85(1):63–81, 1970

    1970

  42. [50]

    Hale and Sjoerd M

    Jack K. Hale and Sjoerd M. Verduyn Lunel.Introduction to Functional Differential Equations, volume 99. Springer Science & Business Media, 1993

    1993

  43. [51]

    A neural RDE approach for continuous- time non-Markovian stochastic control problems

    Melker H¨ oglund, Emilio Ferrucci, Camilo Hern´ andez, Aitor Muguruza Gonzalez, Cristopher Salvi, Leandro S´ anchez-Betancourt, and Yufei Zhang. A neural RDE approach for continuous- time non-Markovian stochastic control problems. InICML 2023 Workshop on New Frontiers in Learn...

    2023

  44. [52]

    Evelyn Hutchinson

    G. Evelyn Hutchinson. Circular causal systems in ecology.Ann. NY Acad. Sci, 50(4):221–246, 1948

    1948

  45. [53]

    Generalized ordinary differential equations in Banach spaces and applications to functional equations.Bol

    Carlos Imaz and Zdenek Vorel. Generalized ordinary differential equations in Banach spaces and applications to functional equations.Bol. Soc. Mat. Mexicana, 11:47–59, 1966

    1966

  46. [54]

    A uniformly differentiable approximation scheme for delay systems using splines.Applied Mathematics and Optimization, 23(3):217–262, 1991

    Kazufumi Ito and Franz Kappel. A uniformly differentiable approximation scheme for delay systems using splines.Applied Mathematics and Optimization, 23(3):217–262, 1991

    1991

  47. [55]

    Stephen Jones

    G. Stephen Jones. Hereditary structure in differential equations.Mathematical systems theory, 1(3):263–278, 1967

    1967

  48. [56]

    Michael I. Jordan. Serial order: A parallel distributed processing approach. Technical Report ICS Report 8604, Institute for Cognitive Science, University of California, San Diego, La Jolla, CA, 1986. 57

    1986

  49. [57]

    Sussmann

    Velimir Jurdjevic and Hector J. Sussmann. Control systems on Lie groups.Journal of Differential Equations, 12(2):313–329, 1972

    1972

  50. [58]

    Littman, and Anthony R

    Leslie Pack Kaelbling, Michael L. Littman, and Anthony R. Cassandra. Planning and acting in partially observable stochastic domains.Artificial Intelligence, 101(1–2):99–134, 1998

    1998

  51. [59]

    Rudolf E. Kalman. A new approach to linear filtering and prediction problems.Journal of Basic Engineering, 82(1):35–45, 1960

    1960

  52. [60]

    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

  53. [61]

    Spline approximation for retarded systems and the Riccati equation.SIAM Journal on Control and Optimization, 25(4):1082–1117, 1987

    Franz Kappel and Dietmar Salamon. Spline approximation for retarded systems and the Riccati equation.SIAM Journal on Control and Optimization, 25(4):1082–1117, 1987

    1987

  54. [62]

    On the stability properties of spline approximations for retarded systems.SIAM Journal on Control and Optimization, 27(2):407–431, 1989

    Franz Kappel and Dietmar Salamon. On the stability properties of spline approximations for retarded systems.SIAM Journal on Control and Optimization, 27(2):407–431, 1989

    1989

  55. [63]

    McKendrick

    William Ogilvy Kermack and Anderson G. McKendrick. A contribution to the mathematical theory of epidemics.Proceedings of the royal society of london. Series A, Containing papers of a mathematical and physical character, 115(772):700–721, 1927

    1927

  56. [64]

    Neural controlled differential equations for irregular time series

    Patrick Kidger, James Morrill, James Foster, and Terry Lyons. Neural controlled differential equations for irregular time series. InAdvances in Neural Information Processing Systems, volume 33, pages 6696–6707. Curran Associates, Inc., 2020

    2020

  57. [65]

    I., volume 159 ofGrundlehren Math

    Gottfried K¨ othe.Topological Vector Spaces. I., volume 159 ofGrundlehren Math. Wiss. Springer, Cham, 1969

    1969

  58. [66]

    Optimal estimation of generic dynamics by path-dependent neural jump ODEs.arXiv Preprint: 2206.14284, 2022

    Florian Krach, Marc N¨ ubel, and Josef Teichmann. Optimal estimation of generic dynamics by path-dependent neural jump ODEs.arXiv Preprint: 2206.14284, 2022

    2022 arXiv

  59. [67]

    Kwossek, Andreas Neuenkirch, and David J

    Anna P. Kwossek, Andreas Neuenkirch, and David J. Pr¨ omel. Functional differential equations driven by c` adl` ag rough paths.Electron. J. Probab., 30:32, 2025. Id/No 117

    2025

  60. [68]

    J. LaSalle. Uniqueness theorems and successive approximations.Annals of Mathematics, 50(3):722–730, 1949

    1949

  61. [69]

    Space of signatures as inverse limits of Carnot groups

    Enrico Le Donne and Roger Z¨ ust. Space of signatures as inverse limits of Carnot groups. ESAIM: Control, Optimisation and Calculus of Variations, 27:37, 2021

    2021

  62. [70]

    Lee.Introduction to Smooth Manifolds, volume 218 ofGraduate Texts in Mathe- matics

    John M. Lee.Introduction to Smooth Manifolds, volume 218 ofGraduate Texts in Mathe- matics. Springer, 2nd edition, 2013

    2013

  63. [71]

    Oxford University Press, 2002

    Terry Lyons and Zhongmin Qian.System Control and Rough Paths. Oxford University Press, 2002

    2002

  64. [72]

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

    Terry J Lyons. Differential equations driven by rough signals.Revista Matem´ atica Iberoamer- icana, 14(2):215–310, 1998. 58

    1998

  65. [73]

    Lyons, Michael J

    Terry J. Lyons, Michael J. Caruana, and Thierry L´ evy.Differential Equations Driven by Rough Paths: Ecole d’Et´ e de Probabilit´ es de Saint-Flour XXXIV-2004. Springer, 2007

    2004

  66. [74]

    On Lipschitz vector fields and the Cauchy problem in homogeneous groups.Communications in Contemporary Mathematics, 20(05):1750057, 2018

    Valentino Magnani and Dario Trevisan. On Lipschitz vector fields and the Cauchy problem in homogeneous groups.Communications in Contemporary Mathematics, 20(05):1750057, 2018

    2018

  67. [75]

    edition, 2007

    Xuerong Mao.Stochastic differential equations and applications.Chichester: Horwood Publishing, 2nd ed. edition, 2007

    2007

  68. [76]

    R. K. Miller. Linear Volterra integrodifferential equations as semigroups.Funkc. Ekvacioj, Ser. Int., 17:39–55, 1974

    1974

  69. [77]

    Directional stability of automatically steered bodies.Journal of the American Society for Naval Engineers, 34(2):280–309, 1922

    Nicolas Minorsky. Directional stability of automatically steered bodies.Journal of the American Society for Naval Engineers, 34(2):280–309, 1922

    1922

  70. [78]

    S. E. A. Mohammed.Stochastic Functional Differential Equations, volume 99 ofRes. Notes Math., San Franc.Pitman Publishing, London, 1984

    1984

  71. [79]

    Neural rough differential equations for long time series

    James Morrill, Cristopher Salvi, Patrick Kidger, and James Foster. Neural rough differential equations for long time series. In Marina Meila and Tong Zhang, editors,Proceedings of the 38th International Conference on Machine Learning, volume 139 ofProceedings of Machine Learni...

    2021

  72. [80]

    Munkres.Topology.Upper Saddle River, NJ: Prentice Hall, 2nd edition, 2000

    James R. Munkres.Topology.Upper Saddle River, NJ: Prentice Hall, 2nd edition, 2000

    2000

  73. [81]

    L. W. Neustadt. On the solutions of certain integral-like operator equations. Existence, uniqueness and dependence theorems.Arch. Ration. Mech. Anal., 38:131–160, 1970

    1970

  74. [82]

    Parameter estimation for rough differential equations.Ann

    Anastasia Papavasiliou and Christophe Ladroue. Parameter estimation for rough differential equations.Ann. Stat., 39(4):2047–2073, 2011

  75. [83]

    Learning non-Markovian dynamical systems with signature-based encoders

    Eliott Pradeleix, R´ emy Hosseinkhan-Boucher, Alena Shilova, Onofrio Semeraro, and Lionel Mathelin. Learning non-Markovian dynamical systems with signature-based encoders. In Cec ´ ılia Coelho, Bernd Zimmering, M. Fernanda P. Costa, Lu ´ ıs L. Ferr´ as, and Oliver Niggemann, e...

    2025

  76. [84]

    Protter.Stochastic Integration and Differential Equations, volume 21 ofAppl

    Philip E. Protter.Stochastic Integration and Differential Equations, volume 21 ofAppl. Math. (N. Y.). Berlin: Springer, 2nd ed. edition, 2004

    2004

  77. [85]

    Lie elements and an algebra associated with shuffles.Annals of Mathematics, 68(2):210–220, 1958

    Rimhak Ree. Lie elements and an algebra associated with shuffles.Annals of Mathematics, 68(2):210–220, 1958

    1958

  78. [86]

    Oxford University Press, 1993

    Christophe Reutenauer.Free Lie Algebras. Oxford University Press, 1993. 59

    1993

  79. [87]

    L. C. G. Rogers and David Williams.Diffusions, Markov processes, and martingales. Vol. 2: Itˆ o calculus.Cambridge: Cambridge University Press, 2nd ed. edition, 2000

    2000

  80. [88]

    Der Fixpunktsatz in Funktionalra¨ umen.Studia Mathematica, 2(1):171–180, 1930

    Juliusz Schauder. Der Fixpunktsatz in Funktionalra¨ umen.Studia Mathematica, 2(1):171–180, 1930

    1930

  81. [89]

    Cambridge University Press, 2022

    Alexander Schmeding.An Introduction to Infinite-Dimensional Differential Geometry, volume 202 ofCambridge Studies in Advanced Mathematics. Cambridge University Press, 2022

    2022

  82. [90]

    Path-dependent SDEs: solutions and parameter estimation.arXiv preprint arXiv:2505.22646, 2025

    Pardis Semnani, Vincent Guan, Elina Robeva, and Darrick Lee. Path-dependent SDEs: solutions and parameter estimation.arXiv preprint arXiv:2505.22646, 2025

    2025 arXiv

  83. [91]

    Sussmann

    Hector J. Sussmann. Lie brackets and local controllability: A sufficient condition for scalar-input systems.SIAM Journal on Control and Optimization, 21(5):686–713, 1983

    1983

  84. [92]

    Sutton and Andrew G

    Richard S. Sutton and Andrew G. Barto.Reinforcement Learning: An Introduction. MIT Press, Cambridge, MA, 2 edition, 2018

    2018

  85. [93]

    Cambridge University Press, 2022

    Jan van Neerven.Functional Analysis, volume 201. Cambridge University Press, 2022

    2022

  86. [94]

    Sulle equazioni integro-differenziali della theoria dell’elasticit` a.Atti Reale Accad

    Vito Volterra. Sulle equazioni integro-differenziali della theoria dell’elasticit` a.Atti Reale Accad. naz. Lincei. Rend. Cl. sci. fis., mat. e natur., 18:295–300, 1909

    1909

  87. [95]

    Gauthier-Villars, 1913

    Vito Volterra.Le¸ cons sur les ´ equations int´ egrales et les ´ equations int´ egro-diff´ erentielles: Le¸ cons profess´ ees ` a la Facult´ e des Sciences de Rome en 1910. Gauthier-Villars, 1913

    1910

  88. [96]

    Sur la th´ eorie math´ ematique des ph´ enom` enes h´ er´ editaires.Journal de math´ ematiques pures et appliqu´ ees, 7:249–298, 1928

    Vito Volterra. Sur la th´ eorie math´ ematique des ph´ enom` enes h´ er´ editaires.Journal de math´ ematiques pures et appliqu´ ees, 7:249–298, 1928

    1928

  89. [97]

    Log neural controlled differential equations: The Lie brackets make a difference

    Benjamin Walker, Andrew Donald Mcleod, Tiexin Qin, Yichuan Cheng, Haoliang Li, and Terry Lyons. Log neural controlled differential equations: The Lie brackets make a difference. In Ruslan Salakhutdinov, Zico Kolter, Katherine Heller, Adrian Weller, Nuria Oliver, Jonathan Scarl...

    2024

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.