Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Sparse Signature Coefficient Recovery via Kernels

T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper claims that any single signature coefficient of a path at any depth can be recovered from $O(2^n)$ signature kernel evaluations, each reducible to an $O(L)$ parallel Goursat PDE solve, bypassing the exponential cost of…

desk verdict Genuinely new algebraic filter for sparse signature coefficients; the numerics section needs a real convergence proof before the complexity claim is solid. read the letter →

arxiv 2412.08579 v2 pith:YRMZW2VY submitted 2024-12-11 math.NA cs.NAstat.ML

Daniil Shmelev , Cristopher Salvi This is my paper · ORCID
classification math.NAcs.NAstat.ML MSC 60L1065Y05
keywords pathsignaturessignaturekernelssparsecoefficientrecoveryiteratedintegralsGoursatPDEVandermondesystemsGPUparallelcomputationsequentialdata
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 proposes a way to compute one iterated integral, one signature coefficient, of a path without computing the whole truncated signature. The idea is to build a filter in signature space whose inner product with a path's signature is exactly the desired coefficient, and then to express that inner product as a linear combination of signature kernels. Because each signature kernel solves a Goursat PDE, every kernel evaluation can be done in parallel in $O(L)$ time, so the depth of the coefficient stops being a bottleneck: the full recovery runs in $O(L)$ parallel time at the price of $O(2^n)$ kernel evaluations. This matters because signature coefficients are the feature basis for many machine-learning methods on sequential data, and existing dense or single-coefficient algorithms scale polynomially with depth.

What carries the argument

The load-bearing object is a filter in signature space, $F = \sum_{i=0}^M \alpha_i D^{(n)}_{\lambda} S(\beta_i \lambda \odot z)\big|_{\lambda=0}$, paired with the identity that $\langle S(x_I), S(y) \rangle$ is the signature kernel $k_{x_I, y}$. The axis path $z = (e_1 * e_2 * \cdots * e_n)_t$ has signature $S(z)_J = 1$ when $J = (1, \ldots, n)$ and $0$ for the other permutations of $(1, \ldots, n)$, so it isolates a single ordering of the letters; the derivative $D^{(n)}_{\lambda}$ removes every multi-index whose letters are not a permutation of $I$. A generalized Vandermonde system $B_{n,M} \alpha = e_1$ selects the $\alpha_i, \beta_i$ so that the combined filter keeps level $n$ intact and annihilates levels $n+1$ through $n+M$, while the factorial decay of signature levels bounds the remaining tail. The whole construction is efficient because the signature kernel solves a Goursat PDE, for which the paper adapts the parallel finite-difference scheme (4.3).

What would settle it

Fix a piecewise-linear path $x$ and a multi-index $I$, compute the reference value $S(x)_I$ by direct iterated integration or a high-truncation Chen computation, and compare it with the kernel-filter recovery at increasing dyadic orders $\gamma_1 = \gamma_2$ of scheme (4.3). The paper's claim predicts the difference decays like $O(2^{-\gamma})$ and approaches zero; a plateau, or a limit different from $S(x)_I$, would show that the numerical premise of Theorem 4.1 does not hold for the axis-path case.

Watch

Extended reading notes

Core claim

The central result is Theorem 3.2: for any $x \in C^1([0,1], \mathbb{R}^d)$ and multi-index $I$, there exist scalars $\alpha_i(M)$ and $\beta_i(M)$ such that the sum $\sum_{i=0}^M \alpha_i D^{(n)}_{\lambda} k^{\lambda-\beta_i}_{x_I, z}\big|_{\lambda=0}$ converges to $S(x)_I$ as $M \to \infty$, where $z$ is the axis path $e_1 * e_2 * \cdots * e_n$. The operator $D^{(n)}_{\lambda}$ is a finite difference chosen to be exact on linear functions, and the Vandermonde choice of $\alpha_i, \beta_i$ zeros signature levels $n+1, \ldots, n+M$, leaving only the factorial-decaying tail as error. The kernel $k$ solves the Goursat PDE $\partial^2 k/\partial t \partial s = \langle \dot{x}_t, \dot{y}_s \rangle k$, and for the axis path the coefficient reduces to a single piecewise-constant channel of $x$, which is what makes the $O(L)$ parallel evaluation possible. The same filter construction generalizes to block sums of permuted coefficients (Theorem 3.4), and with truncated kernels the anagram-class recovery is exact (Proposition 2.4).

Load-bearing premise

The whole procedure inherits its accuracy from the way the kernel equation is solved numerically, and the paper assumes, without proof, that the existing error bound applies to the modified numerical scheme used for the axis-path case; if that assumption fails, the recovered coefficient carries an unknown error.

Editorial extensions

If this is right

  • A single signature coefficient at depth $n$ can be computed in $O(L)$ parallel time (serial cost $O(L n 2^n)$), so on GPU hardware the depth of the needed feature no longer multiplies the runtime the way it does with Chen's relation.
  • Any evaluation of a sparse linear functional on the signature, such as a sparse regression model over path features, can be restricted to exactly the coefficients with nonzero weight and thereby skip the exponential dense tensor computation entirely.
  • When one coefficient is computed on the refined PDE grid, the stored grid contains the kernels for all its initial-segment sub-coefficients (and, with forward differences, all sub-coefficients), so whole families of related coefficients come out of the same computation at no extra kernel cost.
  • For sparse CDEs whose directional graph has maximum out-degree $\Delta_+$, the $N$-step Euler scheme needs only $O(\Delta_+^N)$ signature coefficients rather than $O(d^N)$, and for regular graphs the scheme sparsity approaches $m^{1-N}$, keeping high-order Euler schemes practical.

Reading between the lines

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

  • If the paper's recovery claim is right, the same filter idea could become a general sparse-signature query primitive: any coefficient or linear combination that one can encode as the support of an axis-like path could be recovered by the same kernel expansion, possibly by combining this method with signature-inversion techniques to build filters for arbitrary groups of coefficients.
  • Because the axis-path kernel PDE has a piecewise-constant coefficient on each subinterval, one could replace the finite-difference solver with an exact stepwise computation using the tensor-exponential form of the kernel on each cell, which would remove the discretization error that the paper currently inherits from the numerical scheme.
  • The paper fixes the finite-difference step size and uses small scaling depths ($M = 1$ or $2$) in experiments; the error bound in Proposition 2.8 suggests an adaptive strategy that chooses step size and $M$ per coefficient could reduce the number of kernel evaluations, especially for deep coefficients where factorial decay already suppresses the tail.
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

3 major / 6 minor

Summary. The paper proposes a method for recovering a single signature coefficient S(x)_I at arbitrary depth n without computing the full truncated signature. The construction is algebraic: a filter in the span of signatures is built by combining finite-difference approximations to mixed derivatives with respect to component-wise path scalings, and the inner product of the signature with this filter is evaluated as a linear combination of signature kernels. The kernels are computed by solving a Goursat PDE, and the paper claims a serial complexity of O(L n 2^n) and a parallel complexity of O(L). The main theoretical result is Theorem 3.2, which states that a Vandermonde-weighted finite-difference combination of kernels with an axis path converges to S(x)_I as the scaling depth M tends to infinity. The paper also includes error bounds, experiments on random paths, and an application to sparse N-step Euler schemes for CDEs.

Significance. The algebraic construction is elegant and appears correct: the anagram-class and order-isolation propositions are checkable, Theorem 3.2 is plausible and well-supported by the appendices, and the explicit factorial-decay error bound in Proposition 2.8 is a useful falsifiable prediction that the experiments reproduce. If the numerical PDE step is justified, this would be a novel complexity improvement for computing deep isolated signature coefficients, and the paper provides public code. The main weakness is also load-bearing: the displayed numerical scheme (4.3) is not proved to converge, the cited convergence theorem is not shown to apply to this specific stencil, and the scheme as written omits the β_iλ scaling required by Theorem 3.2. The practical claims, including the O(L) parallel complexity, are therefore conditional on a numerical step that the manuscript does not yet adequately justify.

major comments (3)
  1. [Section 4, Eq. (4.3) and Theorem 4.1] The convergence of the numerical scheme is load-bearing but not established. The text says that the scheme (4.3) is proposed 'taking inspiration from [13,38,40]' and then labels Theorem 4.1 as [38, Theorem 3.5]. Since (4.3) is not shown to be the scheme of [38], the cited convergence theorem does not automatically apply, and no proof is given for the modified stencil or for the special piecewise-constant coefficient in (4.2). More seriously, (4.3) contains no factor of β_iλ_j, whereas Theorem 3.2 requires kernels of the form k_{x_I, β_i λ⊙z}; as written, the scheme solves the PDE for the unscaled axis path z. Please either prove convergence for (4.3), or use exactly the scheme from [38] whose convergence is proved, and specify explicitly how β_iλ enters each kernel evaluation. This is necessary to support the claimed O(L) parallel complexity.
  2. [Section 4, Eq. (4.2) and definition of z] The parameterization of the axis path z is inconsistent. The definition z_t = (e_1 ∗ ⋯ ∗ e_n)_t for t ∈ [0,1], where each (e_i)_t is the linear path from 0 to e_i, is not a path on [0,1] under the standard concatenation convention; the PDE is then stated with s ∈ [m/d, (m+1)/d] and 'node at i/d', while the grid P_{γ1,γ2} uses j up to 2^{γ2} n. Depending on the intended convention, the segment speed n appears in the PDE coefficient, and the claimed scheme (4.3) may not correspond to the same path as Theorem 3.2. Please fix the definition of z, the interval for s, and the grid so that (4.2)–(4.3) describe the same path used in the algebraic results.
  3. [Section 2, Notation 2.3 and Proposition 2.4] The exactness condition on D^{(n)}_λ is misstated. For the forward and central mixed differences used in the paper, the relevant property is exactness on the monomial λ_1⋯λ_n, together with the annihilation of monomials with repeated or missing variables, at the level-n contribution. The statement that D^{(n)}_λ is 'exact on linear functions of λ_1,...,λ_n' is not the needed condition, and a difference operator exact only on affine functions would not satisfy equation (2.6). Proposition 2.4 currently says the result follows immediately, but the verification that D^{(n)}_λ((λ_1+⋯+λ_n)^n) = n! is a nontrivial combinatorial fact. Please correct the definition and include this verification.
minor comments (6)
  1. [Section 3, paragraph before Proposition 3.1] For the linear path y_t = te, the signature coefficient is S(y)_J = 1/|J|!, not |J|!; the sentence claiming that 'S(y)_J = |J|!' extracts the anagram class weighted by n! is incorrect as written. The displayed component-wise limit in (3.1) also appears to describe the filter based on z rather than S(β_iλ⊙x_I), whose limit would carry the coefficients S(x_I)_J.
  2. [Section 3, proof of Proposition 3.1] The proof of Proposition 3.1 cites Proposition 2.6, but the argument is that of Proposition 2.1; please correct the cross-reference.
  3. [Section 4, Theorem 4.1] If Theorem 4.1 is intended as a verbatim citation of [38, Theorem 3.5], the paper should state explicitly that (4.3) is the scheme from [38]; otherwise the label 'Theorem 4.1 ([38, Theorem 3.5])' is misleading.
  4. [Section 4.1] The inclusion K_J ⊂ K_I is asserted without specifying the identification of grid points, time intervals, and the extension of the central-difference nodes; a short explicit mapping would make the batch-retrieval claim easier to verify.
  5. [Throughout] There are several small typos and notational inconsistencies: 'singular coefficient' in the abstract should be 'single coefficient'; the title line contains spacing artefacts; and in (4.2), 't = i/d' should be a statement about s, with the convention for the segment length made consistent.
  6. [Section 2, Proposition 2.6 and Example 2.7] In the condition (2.9) and the uniform-choice calculation, the paper should specify the constant C that makes the displayed O(M^{n+1}2^M) bound compatible with (2.9); the current text leaves the choice implicit.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the recovered coefficient is produced by an explicit, x-independent filter; the only load-bearing citation is a published PDE theorem whose applicability to the modified stencil is a numerical proof gap, not a circular reduction.

full rationale

The derivation of Theorem 3.2 is self-contained and constructive rather than circular. The filter is not fitted to the target coefficient: z is the fixed axis path e1*...*en, alpha_i are the explicit inverse-Vandermonde coefficients of Proposition 2.5, beta_i are chosen independently of x (e.g., uniform or nth-root scalings), and D_lambda^(n) is a fixed finite-difference operator. The proof expands k_{x_I, beta_i lambda circle z} level by level, observes that mixed differentiation at lambda=0 retains only the level-n squarefree monomial, that the Vandermonde system (2.8) zeroes levels n+1 through n+M, and that factorial decay controls the remaining tail; the axis path's signature S(z)_J = 1_{J=(1,...,n)} is verified from Chen's relation, not assumed. Thus S(x)_I appears on both sides only as the target of an explicit projection, not as a fitted quantity or as an input to the construction. The only load-bearing external input is the signature-kernel Goursat PDE (Theorem 1.12) and the convergence bound (Theorem 4.1), both cited from [38]; [38] is a published, parameter-free theorem whose assumptions do not include the target coefficient, so citing it does not make the argument circular. Caveat, flagged for correctness rather than circularity: Theorem 4.1 is stated as a citation without demonstrating that the dyadically modified stencil (4.3) satisfies the hypotheses of [38, Theorem 3.5] for the axis-path PDE (4.2); this is an omitted numerical proof, not a reduction of the prediction to its inputs.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

No parameters are fitted to the target coefficient; M, beta_i, h, and gamma are algorithm hyperparameters. The mathematical results rely on standard signature theory, the Goursat PDE theorem from [38], and an unproven adaptation of the PDE convergence theorem to the modified scheme. No unexplained physical or mathematical entities are introduced.

free parameters (4)
  • M (scaling depth) = 1 or 2 in experiments
    Controls the number of Vandermonde kernel terms and the residual error from levels beyond n+M; not fitted to the recovered coefficient, but chosen by the user and required to tend to infinity for exact convergence.
  • beta_i (Vandermonde nodes) = e.g. (i+1)/(M+1) or [(i+1)/(M+1)]^(1/n)
    Distinct nodes in (0,1] satisfying Condition 2.9; the choice affects numerical conditioning and the growth of alpha_i, but not the mathematical limit.
  • h (finite difference step) = 1
    Step size for forward or central mixed differences; exactness on multilinear monomials holds for any h, and h=1 is chosen for numerical stability.
  • gamma1, gamma2 (dyadic refinement orders) = 3 in experiments
    PDE grid refinement parameters that control the discretization error O(2^{-gamma1-gamma2}); they are numerical hyperparameters, not fitted to data.
assumptions (6)
  • domain assumption The signature kernel satisfies the Goursat PDE (Theorem 1.12)
    Core computational tool, cited from [38]; the paper uses it without reproving it.
  • domain assumption The numerical scheme (4.3) converges with error O(2^{-gamma1-gamma2}) for the axis-path kernel PDE (4.2)
    Cited to [38, Theorem 3.5] but not proven for the modified scheme; this is the weakest load-bearing numerical premise.
  • standard math Factorial decay of signature coefficients (Lemma 1.5)
    Used to bound residual contributions from levels deeper than n+M in Propositions 2.6 and 2.8.
  • standard math Finite difference D^(n) is exact on multilinear monomials with each variable exponent at most 1
    Underlies anagram and order isolation; the paper states this incorrectly as exact on linear functions.
  • standard math Interchange of differentiation and infinite summation in the proof of Proposition 2.1
    Justified by uniform convergence of the derivative series; the paper mentions this but does not give details.
  • standard math Vandermonde matrix invertibility and explicit inverse (Proposition 2.5)
    Used to construct coefficients alpha_i that zero levels n+1 through n+M; proven in Appendix B via [1].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Sparse Signature Coefficient Recovery via Kernels." pith.science (2026). https://pith.science/paper/YRMZW2VY

@misc{pith2026241208579,
  author       = {Pith},
  title        = {Pith review of: Sparse Signature Coefficient Recovery via Kernels},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YRMZW2VY}},
  note         = {Machine review of arXiv:2412.08579}
}
read the original abstract

Central to rough path theory is the signature transform of a path, an infinite series of tensors given by the iterated integrals of the underlying path. The signature poses an effective way to capture sequentially ordered information, thanks both to its rich analytic and algebraic properties as well as its universality when used as a basis to approximate functions on path space. Whilst a truncated version of the signature can be efficiently computed using Chen's identity, there is a lack of efficient methods for computing a sparse collection of iterated integrals contained in high levels of the signature. We address this problem by leveraging signature kernels, defined as the inner product of two signatures, and computable efficiently by means of PDE-based methods. By forming a filter in signature space with which to take kernels, one can effectively isolate specific groups of signature coefficients and, in particular, a singular coefficient at any depth of the transform. We show that such a filter can be expressed as a linear combination of suitable signature transforms and demonstrate empirically the effectiveness of our approach. To conclude, we give an example use case for sparse collections of signature coefficients based on the construction of N-step Euler schemes for sparse CDEs.

Figures

Figures reproduced from arXiv: 2412.08579 by the authors.

Figure 1
Figure 1. Average errors when computing S(x) (1,...,n) over 1,000 random paths constrained to [0, 1]d , with path length L = 150 and scaling depth M = 3. We compute average absolute errors (left) and average absolute errors scaled by average coefficient magnitude (right) [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Average errors when computing S(x) (1,2,...,n) for random paths in [0, 1]d , with L = 100 (top), L = 500 (middle) and L = 1000 (bottom). Dyadic order of the PDE solver is set to 3 [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Dependence of runtime on coefficient depth, using serial versus parallel kernel computations [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Generational model with n = 13 approximated using a 5-step Euler scheme, leading to a sparsity of s(E 5 s,t) = 5 × 10−5 . Model parameters are chosen as a(k) = 1, b(k) = f(k) = 10, c(k) = k/14, d(k) = (14 − k)/14. Epidemic is chosen to never occur (top left), start lat…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Advances in Neural Controlled Differential Equations

    cs.LG 2026-07 conditional novelty 7.0 of 10

    Linear NCDEs replace non-linear vector fields with linear ones, enabling parallel-in-time training via associative scans while retaining maximal theoretical expressivity and achieving state-of-the-art time series perf...

Reference graph

Works this paper leans on

39 extracted references · 27 canonical work pages · cited by 1 Pith paper

  1. [38]

    Salvi, T

    C. Salvi, T. Cass, J. Foster, T. Lyons, and W. Yang , The signature kernel is the solution of a Goursat PDE , SIAM Journal on Mathematics of Data Science, 3 (2021), pp. 873–899

    work page 2021

  2. [1]

    Arafat and M

    A. Arafat and M. El-Mikkawy , A fast novel recursive algorithm for computing the inverse of a generalized van- dermonde matrix , Axioms, 12 (2022), p. 27

    work page 2022

  3. [2]

    I. P. Arribas, C. Salvi, and L. Szpruch, Sig-sdes model for quantitative finance , in ACM International Conference on AI in Finance, 2020

    work page 2020

  4. [3]

    Barancikova, Z

    B. Barancikova, Z. Huang, and C. Salvi , Sigdiffusions: Score-based diffusion models for long time series via log-signature embeddings, arXiv preprint arXiv:2406.10354, (2024)

    arXiv 2024

  5. [4]

    Boedihardjo, X

    H. Boedihardjo, X. Geng, T. Lyons, and D. Yang , The signature of a rough path: uniqueness , Advances in Mathematics, 293 (2016), pp. 720–737

    work page 2016

  6. [5]

    Cass and C

    T. Cass and C. Salvi , Lecture notes on rough paths and applications to machine learning , arXiv preprint arXiv:2404.06583, (2024)

    arXiv 2024

  7. [6]

    Cass and W

    T. Cass and W. F. Turner , Topologies on unparameterised path space, Journal of Functional Analysis, 286 (2024), p. 110261

    work page 2024

  8. [7]

    Chang and T

    J. Chang and T. Lyons, Insertion algorithm for inverting the signature of a path , arXiv preprint arXiv:1907.08423, (2019)

    arXiv 2019

Show all 39 references
  1. [8]

    Chen, Iterated integrals and exponential homomorphisms, Proceedings of the London Mathematical Society, 3 (1954), pp

    K.-T. Chen, Iterated integrals and exponential homomorphisms, Proceedings of the London Mathematical Society, 3 (1954), pp. 502–512

    1954

  2. [9]

    Chen, Integration of paths, geometric invariants and a generalized baker-hausdorff formula , Annals of Math- ematics, (1957), pp

    K.-T. Chen, Integration of paths, geometric invariants and a generalized baker-hausdorff formula , Annals of Math- ematics, (1957), pp. 163–178

    1957

  3. [10]

    N. M. Cirone, M. Lemercier, and C. Salvi , Neural signature kernels as infinite-width-depth-limits of controlled resnets, arXiv preprint arXiv:2303.17671, (2023)

    2023 arXiv

  4. [11]

    N. M. Cirone, A. Orvieto, B. W alker, C. Salvi, and T. Lyons , Theoretical foundations of deep selective state- space models, arXiv preprint arXiv:2402.19047, (2024)

    2024 arXiv

  5. [12]

    Cochrane, P

    T. Cochrane, P. Foster, V. Chhabra, M. Lemercier, T. Lyons, and C. Salvi , Sk-tree: a systematic malware detection algorithm on streaming trees via the signature kernel , in 2021 IEEE International Conference on Cyber Security and Resilience (CSR), IEEE, 2021, pp. 35–40

    2021

  6. [13]

    J. T. Day, A runge-kutta method for the numerical solution of the goursat problem in hyperbolic partial differential equations., Comput. J., 9 (1966), pp. 81–83

    1966

  7. [14]

    T. L. et al , Coropa computational rough paths (software library) , (2010), http://coropa.sourceforge.net/

    2010

  8. [15]

    Fermanian, T

    A. Fermanian, T. Lyons, J. Morrill, and C. Salvi, New directions in the applications of rough path theory , IEEE BITS the Information Theory Magazine, (2023)

    2023

  9. [16]

    P. K. Friz and N. B. Victoir , Multidimensional stochastic processes as rough paths: theory and applications , vol. 120, Cambridge University Press, 2010

    2010

  10. [17]

    Hambly and T

    B. Hambly and T. Lyons, Uniqueness for the signature of a path of bounded variation and the reduced path group , Annals of Mathematics, (2010), pp. 109–167

    2010

  11. [18]

    Hoglund, E

    M. Hoglund, E. Ferrucci, C. Hern ´andez, A. M. Gonzalez, C. Salvi, L. Sanchez-Betancourt, and Y. Zhang, A neural rde approach for continuous-time non-markovian stochastic control problems, arXiv preprint arXiv:2306.14258, (2023). SPARSE SIGNATURE COEFFICIENT RECOVERY VIA KERNELS 23

    2023 arXiv

  12. [19]

    Holberg and C

    C. Holberg and C. Salvi , Exact gradients for stochastic spiking neural networks driven by rough signals , arXiv preprint arXiv:2405.13587, (2024)

    2024 arXiv

  13. [20]

    Horvath, M

    B. Horvath, M. Lemercier, C. Liu, T. Lyons, and C. Salvi, Optimal stopping via distribution regression: a higher rank signature approach, arXiv preprint arXiv:2304.01479, (2023)

    2023 arXiv

  14. [21]

    Z. Issa, B. Horvath, M. Lemercier, and C. Salvi , Non-adversarial training of neural sdes with signature kernel scores, Advances in Neural Information Processing Systems, 36 (2024)

    2024

  15. [22]

    Kidger, P

    P. Kidger, P. Bonnier, I. Perez Arribas, C. Salvi, and T. Lyons, Deep signature transforms, Advances in Neural Information Processing Systems, 32 (2019)

    2019

  16. [24]

    Kidger and T

    P. Kidger and T. Lyons , Signatory: differentiable computations of the signature and logsignature transforms, on both CPU and GPU , arXiv preprint arXiv:2001.00706, (2020)

    2020 arXiv

  17. [25]

    F. J. Kir ´aly and H. Oberhauser, Kernels for sequentially ordered data , Journal of Machine Learning Research, 20 (2019), pp. 1–45

    2019

  18. [26]

    Lees, The goursat problem, Journal of the Society for Industrial and Applied Mathematics, 8 (1960), pp

    M. Lees, The goursat problem, Journal of the Society for Industrial and Applied Mathematics, 8 (1960), pp. 518–530

    1960

  19. [27]

    Lemercier, C

    M. Lemercier, C. Salvi, T. Cass, E. V. Bonilla, T. Damoulas, and T. Lyons , Siggpde: Scaling sparse gaussian processes on sequential data , in International Conference on Machine Learning, PMLR, 2021

    2021

  20. [28]

    Lemercier, C

    M. Lemercier, C. Salvi, T. Damoulas, E. Bonilla, and T. Lyons , Distribution regression for sequential data , in International Conference on Artificial Intelligence and Statistics, PMLR, 2021, pp. 3754–3762

    2021

  21. [29]

    Lyons, Rough paths, signatures and the modelling of functions on streams , International Congress of Mathema- ticians, Seoul, (2014)

    T. Lyons, Rough paths, signatures and the modelling of functions on streams , International Congress of Mathema- ticians, Seoul, (2014)

    2014

  22. [30]

    Lyons, M

    T. Lyons, M. Caruana, and T. L ´evy, Differential equations driven by rough paths , Ecole d’´ et´ e de Probabilit´ es de Saint-Flour XXXIV, (2004), pp. 1–93

    2004

  23. [31]

    Lyons and A

    T. Lyons and A. D. McLeod , Signature methods in machine learning , arXiv preprint arXiv:2206.14674, (2022)

    2022 arXiv

  24. [32]

    T. J. Lyons and W. Xu , Hyperbolic development and inversion of signature , Journal of Functional Analysis, 272 (2017), pp. 2933–2955

    2017

  25. [33]

    T. J. Lyons and W. Xu, Inverting the signature of a path , Journal of the European Mathematical Society, 20 (2018), pp. 1655–1687

    2018

  26. [34]

    Manten, C

    G. Manten, C. Casolo, E. Ferrucci, S. W. Mogensen, C. Salvi, and N. Kilbertus, Signature kernel conditional independence tests in causal discovery for stochastic processes , arXiv preprint arXiv:2402.18477, (2024)

    2024 arXiv

  27. [35]

    Morrill, C

    J. Morrill, C. Salvi, P. Kidger, and J. Foster , Neural rough differential equations for long time series , in International Conference on Machine Learning, PMLR, 2021, pp. 7829–7838

    2021

  28. [36]

    Pannier and C

    A. Pannier and C. Salvi, A path-dependent pde solver based on signature kernels , arXiv preprint arXiv:2403.11738, (2024)

    2024

  29. [37]

    Reizenstein and B

    J. Reizenstein and B. Graham , The iisignature library: efficient calculation of iterated-integral signatures and log signatures, arXiv preprint arXiv:1802.08252, (2018)

    2018 arXiv

  30. [39]

    Salvi, M

    C. Salvi, M. Lemercier, C. Liu, B. Horvath, T. Damoulas, and T. Lyons, Higher order kernel mean embeddings to capture filtrations of stochastic processes , Advances in Neural Information Processing Systems, 34 (2021), pp. 16635–16647

    2021

  31. [40]

    W azwaz, On the numerical solution of the goursat problem , Applied mathematics and computation, 59 (1993), pp

    A.-M. W azwaz, On the numerical solution of the goursat problem , Applied mathematics and computation, 59 (1993), pp. 89–95

    1993

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.