pith. sign in

arxiv: 2605.18406 · v1 · pith:CZ23SYPMnew · submitted 2026-05-18 · 🧮 math.NA · cs.NA· stat.ML

Computational aspects of the Volterra Signature

Paul P. Hager , Fabian N. Harang , Luca Pelizzari , Samy Tindel This is my paper

Pith reviewed 2026-05-19 23:55 UTC · model grok-4.3

classification 🧮 math.NA cs.NAstat.ML
keywords Volterra signaturepath signatureiterated integralsmatrix-valued kernelsVolterra equationscomputational complexityFFT accelerationstate-space representation
0
0 comments X p. Extension
pith:CZ23SYPM Add to your LaTeX paper What is a Pith Number? 
\usepackage{pith}
\pithnumber{CZ23SYPM}

Prints a linked pith:CZ23SYPM badge after your title and writes the identifier into PDF metadata. Compiles on arXiv with no extra files. Learn more

The pith

The Volterra signature with matrix-valued kernels admits efficient computation via quadratic approximation, FFT acceleration, and low-dimensional recursion.

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

The paper shows that the Volterra signature, formed by inserting general matrix-valued kernels into the iterated integrals of the classical path signature, can be computed by breaking the underlying Chen convolution into separate analytic and arithmetic steps. This split supports a baseline algorithm that runs in O(J squared) time for J time steps, an FFT version that drops to O(J log J) on uniform grids, and an exact recursion that runs in O(J R squared) when the kernel has a state-space form of dimension R. The methods keep the usual dependence on path dimension and truncation level N, and kernels written as sums of scalar functions times constant matrices do not raise the leading complexity. A predictor-corrector finite-difference scheme for the signature kernel itself is also derived and all methods are released in open code.

Core claim

By decomposing the Chen-type convolution relation for the Volterra signature into an analytic kernel-integration part and an arithmetic signature part, the components can be obtained through a general quadratic-time scheme, an FFT-based O(J log J) scheme for convolution kernels on uniform grids, an exact O(J R squared) recursion for state-space kernels of dimension R, and a finite-difference predictor-corrector method, while the number of matrix factors in kernels of the form sum k_p(t-s) A_p leaves the asymptotic cost in J and N unchanged.

What carries the argument

Decomposition of the Chen-type convolution relation into analytic and arithmetic parts, which isolates the kernel integration from the combinatorial operations that build the signature components.

If this is right

  • The Volterra signature becomes practical for long time series because its cost grows only quadratically in the number of steps rather than exponentially in truncation level.
  • Convolution kernels on uniform time grids can be handled in near-linear time, making the method suitable for large uniform datasets.
  • Kernels that admit a low-dimensional state-space realization allow exact computation whose cost scales only with the state dimension squared.
  • Kernels expressed as finite sums of scalar functions times fixed matrices incur no extra asymptotic cost in the number of summands.
  • The signature kernel itself can be approximated by a predictor-corrector finite-difference scheme that inherits the same efficiency gains.

Where Pith is reading between the lines

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

  • These algorithms could be combined with existing signature-based machine-learning pipelines to add memory effects without a prohibitive increase in runtime.
  • The same decomposition strategy might apply to other iterated-integral objects, such as those arising in rough-path theory or controlled differential equations.
  • Numerical tests on real high-frequency financial or physiological data could quantify how much additional predictive power the kernel introduces relative to the classical signature.

Load-bearing premise

The decomposition of the Chen-type convolution relation into analytic and arithmetic parts can be performed without introducing errors that propagate into the final signature components for general matrix-valued kernels.

What would settle it

Compute the Volterra signature up to level 3 for a simple two-dimensional path and a rank-one kernel using both the proposed quadratic scheme and direct numerical quadrature of the defining iterated integrals; the two results must agree to within the expected truncation and quadrature tolerance.

Figures

Figures reproduced from arXiv: 2605.18406 by Fabian N. Harang, Luca Pelizzari, Paul P. Hager, Samy Tindel.

Figure 1
Figure 1. Figure 1: Level-wise sample standard deviation of the factorially adjusted signature levels n! πnSig(x (i) ) for the generated sample paths. Computational costs. To allow for a hardware-independent validation of the compu￾tational costs of the proposed algorithms, we use the number of floating-point operations (FLOPs) as the main reference quantity. Since our implementation uses the JAX backend, FLOP counts are obta… view at source ↗
Figure 2
Figure 2. Figure 2: Convergence of the general approximative Volterra signature schemes under dyadic refinement. The plotted quantities are the factorially adjusted level errors δ scheme n,λ . Values in parentheses denote the fitted log– log slope of the error against the dyadic grid size. Left: β = 0.6. Right: β = 0.1. 100 101 102 103 elapsed time per path (ms) 10−7 10−6 10−5 10−4 10−3 10−2 10−1 100 δVλ 0 1 2 3 4 0 1 2 3 4 0… view at source ↗
Figure 3
Figure 3. Figure 3: Error–runtime tradeoff for the predictor–corrector reference scheme and the proposed higher-order Volterra signature schemes. from Algorithm 1, the predicted asymptotic work is Wquad(J, N) = ( J 2mN , q = 1 with the Horner scheme of Algorithm 3, J 2NmN , q > 1 with the shuffle-recursive scheme of Algorithm 2. For the FFT-accelerated implementation from Algorithm 4, in the uniform-grid convolu￾tional settin… view at source ↗
Figure 4
Figure 4. Figure 4: Computational scaling of the general approximative Volterra signature algorithms. Left: compiler-reported FLOP counts for the qua￾dratic triangular recursion plotted against Wquad(J, N, q). Right: compiler￾reported FLOP counts for the FFT-accelerated implementation plotted against WFFT(J, N, q). Dashed lines indicate per-q unit-slope intercepts fits against largest 40 workloads. A.2. Validation of the fini… view at source ↗
Figure 5
Figure 5. Figure 5: Convergence of the dyadically refined Euler scheme towards the exact benchmark Volterra signature for two representative setups. Finally, we analyze the computational cost of the Volterra signature computation ac￾cording to Algorithms 6 and 7. We first consider the state recursion and readout alone, excluding the precomputation of the coefficients appearing in these algorithms; we comment on the cost of th… view at source ↗
Figure 6
Figure 6. Figure 6: Computational scaling of the finite-state-space Volterra signa￾ture computation. Left: compiler-reported FLOP counts against the pre￾dicted work Wq(J, R, N). Right: measured wall-clock time per path against Wq(J, R, N). Dashed lines indicate per-q unit-slope intercepts fits against largest 40 workloads. A.3. Validation of the signature kernel algorithm. We finally validate the finite￾difference scheme from… view at source ↗
Figure 7
Figure 7. Figure 7: Validation of the Volterra signature kernel algorithm. Left: convergence of the naive, exponential integration, and predictor–corrector schemes against the truncated inner-product reference κ ref,N . Right: compiler-reported total FLOP counts plotted against the leading asymp￾totic work J 2R2 [PITH_FULL_IMAGE:figures/full_fig_p066_7.png] view at source ↗
read the original abstract

The Volterra signature extends the classical path signature by incorporating general matrix-valued kernel into its iterated integral structure, yielding a flexible notion of memory for time series. Its components can be viewed as successive Picard iterates of linear controlled Volterra equations, making their exact computation of additional mathematical interest. However, the kernel introduces substantial algorithmic challenges. We provide a resolution by first decomposing the Chen-type convolution relation established in [arXiv:2603.04525] into analytic and arithmetic parts, and then introducing several efficient algorithms: a general approximative scheme with quadratic complexity $O(J^2)$ in the number of time steps $J$, an FFT-based acceleration with complexity $O(J\log J)$ for convolution kernels on uniform grids, and an exact recursion with complexity $O(JR^2)$ for kernels admitting a state-space representation of dimension $R$; retaining standard signature complexity in the path dimension and truncation level $N$. We further show that the number of factors in matrix-valued kernels of the form $K(t,s)=\sum_p k_p(t-s)A_p$ do not increase the asymptotic complexity in $J$ and $N$. Finally, we derive a finite-difference predictor--corrector scheme for the associated Volterra signature kernel. All algorithms are implemented in the publicly available JAX-based package "tensordev".

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript develops computational methods for the Volterra signature, which augments the classical path signature with general matrix-valued kernels. By decomposing the Chen-type convolution relation from arXiv:2603.04525 into analytic and arithmetic parts, the authors derive a general approximative scheme of complexity O(J²), an FFT acceleration of complexity O(J log J) for convolution kernels on uniform grids, an exact recursion of complexity O(J R²) for kernels with state-space dimension R, and a finite-difference predictor-corrector scheme. They further show that kernels of the form K(t,s)=∑_p k_p(t-s)A_p do not increase asymptotic complexity in J or N, and release a public JAX package implementing all methods.

Significance. If the decomposition preserves the algebraic structure of the iterated integrals without uncontrolled error accumulation, the algorithms would provide a substantial advance in the numerical treatment of Volterra signatures for time-series and rough-path applications. The public JAX implementation and the retention of standard signature complexity in path dimension and truncation level N are concrete strengths that support reproducibility and practical use.

major comments (2)
  1. [Abstract and the section introducing the decomposition] The decomposition of the Chen-type convolution relation (referenced from arXiv:2603.04525) into separate analytic and arithmetic parts underpins every complexity claim in the abstract. The manuscript provides no explicit error bound or algebraic verification that this split preserves the Picard-iterate relations for arbitrary (non-scalar, non-convolution) matrix-valued kernels K(t,s); without such a bound, the O(J²), O(J log J) and O(J R²) statements rest on an unverified assumption that could affect faithfulness of the computed signature components.
  2. [Section describing the exact recursion] The exact recursion of complexity O(J R²) is stated to apply to kernels admitting a state-space representation of dimension R. The manuscript should clarify, with a concrete example or inductive argument, how the state-space matrices are propagated through the Volterra signature levels without reintroducing the full matrix-valued kernel at each step.
minor comments (2)
  1. [Abstract] The abstract mentions a finite-difference predictor-corrector scheme; a brief statement of its stability or consistency order would help readers assess its relation to the other three algorithms.
  2. Notation for the truncation level N and the number of time steps J is used throughout; a short table summarizing the complexity of each method in terms of J, N, R and path dimension would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thorough review and constructive feedback on our manuscript. The comments raise important points about the rigor of the decomposition and the clarity of the recursion. We address each major comment below and will incorporate revisions to strengthen the presentation.

read point-by-point responses

  1. Referee: [Abstract and the section introducing the decomposition] The decomposition of the Chen-type convolution relation (referenced from arXiv:2603.04525) into separate analytic and arithmetic parts underpins every complexity claim in the abstract. The manuscript provides no explicit error bound or algebraic verification that this split preserves the Picard-iterate relations for arbitrary (non-scalar, non-convolution) matrix-valued kernels K(t,s); without such a bound, the O(J²), O(J log J) and O(J R²) statements rest on an unverified assumption that could affect faithfulness of the computed signature components.

    Authors: We thank the referee for this observation. The decomposition follows directly from the Chen-type relation in the referenced work by separating the kernel-dependent analytic integration from the subsequent tensor arithmetic. This split is exact with respect to the defining Picard-iterate structure for general matrix-valued kernels; approximation errors arise only from the numerical treatment of the analytic part (e.g., quadrature). In the revised manuscript we will add an explicit algebraic verification together with a short error-propagation argument showing that the iterated-integral relations are preserved up to the controlled approximation error of the chosen scheme. This will make the complexity statements fully rigorous without altering the reported asymptotics. revision: yes

  2. Referee: [Section describing the exact recursion] The exact recursion of complexity O(J R²) is stated to apply to kernels admitting a state-space representation of dimension R. The manuscript should clarify, with a concrete example or inductive argument, how the state-space matrices are propagated through the Volterra signature levels without reintroducing the full matrix-valued kernel at each step.

    Authors: We agree that additional clarification is helpful. The state-space representation allows the kernel action to be replaced by a linear dynamical system whose state is updated at each time step. In the revision we will insert an inductive argument: assuming the signature up to level k is expressed via auxiliary state vectors of dimension R, the update to level k+1 is obtained by integrating the state-space dynamics against the previous signature component, without ever reconstructing the full kernel matrix. A concrete low-dimensional example (scalar exponential kernel realized by a 1-dimensional state space) will be included to illustrate the propagation explicitly. revision: yes

Circularity Check

0 steps flagged

No circularity: algorithms derived from cited relation without reduction to inputs

full rationale

The paper cites the Chen-type convolution relation from prior work [arXiv:2603.04525] and performs a decomposition into analytic and arithmetic parts to derive new algorithms with stated complexities O(J^2), O(J log J), and O(J R^2). No equation or claim in the abstract reduces any 'prediction' or result to a fitted parameter, self-definition, or tautological renaming within this paper's own equations. The complexities follow from standard analysis of the decomposed convolution on the Volterra signature components, which remain defined via Picard iterates independently of the computational schemes. This is a self-contained derivation building on an external mathematical relation without circular equivalence to its inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work rests on the prior Chen-type convolution relation from the cited arXiv preprint and standard properties of iterated integrals and Picard iteration for Volterra equations. No new free parameters, invented entities, or ad-hoc axioms are introduced in the abstract.

axioms (1)
  • domain assumption The Chen-type convolution relation for Volterra signatures holds and can be decomposed into independent analytic and arithmetic components.
    Invoked to enable the separation that underpins all three proposed algorithms.

pith-pipeline@v0.9.0 · 5778 in / 1328 out tokens · 21524 ms · 2026-05-19T23:55:26.531382+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

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

  • IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    We provide a resolution by first decomposing the Chen-type convolution relation established in [13] into analytic and arithmetic parts, and then introducing several efficient algorithms: a general approximative scheme with quadratic complexity O(J²) ... an exact recursion with complexity O(J R²) for kernels admitting a state-space representation of dimension R

  • IndisputableMonolith/Foundation/AlphaCoordinateFixation.lean alpha_pin_under_high_calibration unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    For kernels of exponential-polynomial and periodic type ... we can make use of a state space lift to provide an exact scheme ... costs are proportional to J × R²

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

Reference graph

Works this paper leans on

43 extracted references · 43 canonical work pages · 1 internal anchor

  1. [1]

    Dover Publications, Inc., USA, 1974

    Milton Abramowitz.Handbook of Mathematical Functions, With Formulas, Graphs, and Mathematical Tables. Dover Publications, Inc., USA, 1974. 55

    work page 1974

  2. [2]

    Al-Mohy and Bahar Arslan

    Awad H. Al-Mohy and Bahar Arslan. The complex step approximation to the higher order Fréchet derivatives of a matrix function.Numer. Algorithms, 87(3):1061–1074, 2021. 37

    work page 2021

  3. [3]

    Springer Finance

    Christian Bayer, Gonçalo dos Reis, Blanka Horvath, and Harald Oberhauser, editors.Signature Meth- ods in Finance: An Introduction with Computational Applications. Springer Finance. Springer, Cham,

    work page

  4. [4]

    2 COMPUTATIONAL ASPECTS OF THE VOLTERRA SIGNATURE 58

    eBook published 07 Nov 2025;©2026 Springer Nature. 2 COMPUTATIONAL ASPECTS OF THE VOLTERRA SIGNATURE 58

    work page 2025

  5. [5]

    Cohen, Terry Lyons, Joël Mouterde, and Benjamin Walker

    Alexandre Bloch, Samuel N. Cohen, Terry Lyons, Joël Mouterde, and Benjamin Walker. The expo- nentially weighted signature, 2026. 3

    work page 2026

  6. [6]

    JAX: composable transformations of Python+NumPy programs, 2018

    James Bradbury, Roy Frostig, Peter Hawkins, Matthew James Johnson, Yash Katariya, Chris Leary, Dougal Maclaurin, George Necula, Adam Paszke, Jake VanderPlas, Skye Wanderman-Milne, and Qiao Zhang. JAX: composable transformations of Python+NumPy programs, 2018. 5

    work page 2018

  7. [7]

    Numerical schemes for signature kernels.SIAM Journal on Numerical Analysis, 63(6), 2025

    Thomas Cass, Francesco Piatti, and Jeffrey Pei. Numerical schemes for signature kernels.SIAM Journal on Numerical Analysis, 63(6), 2025. 4

    work page 2025

  8. [8]

    Integration of paths, geometric invariants and a Generalized Baker–Hausdorff formula

    Kuo-Tsai Chen. Integration of paths, geometric invariants and a Generalized Baker–Hausdorff formula. Annals of Mathematics, 65(1):163–178, 1957. 2

    work page 1957

  9. [9]

    A primer on the signature method in machine learning, 2016

    Ilya Chevyrev and Andrey Kormilitzin. A primer on the signature method in machine learning, 2016. 2

    work page 2016

  10. [10]

    Cooley and John W

    James W. Cooley and John W. Tukey. An algorithm for the machine calculation of complex fourier series.Mathematics of Computation, 19(90):297–301, 1965. 24, 54

    work page 1965

  11. [11]

    Time warping invariants of multidimen- sional time series.Acta Appl

    Joscha Diehl, Kurusch Ebrahimi-Fard, and Nikolas Tapia. Time warping invariants of multidimen- sional time series.Acta Appl. Math., 170:265–290, 2020. 4

    work page 2020

  12. [12]

    Fruits: feature extraction using iterated sums for time series classi- fication.Data Mining and Knowledge Discovery, 38:4122–4156, 2024

    Joscha Diehl and Richard Krieg. Fruits: feature extraction using iterated sums for time series classi- fication.Data Mining and Knowledge Discovery, 38:4122–4156, 2024. 11

    work page 2024

  13. [13]

    Ford, and Alan D

    Kai Diethelm, Neville J. Ford, and Alan D. Freed. Detailed error analysis for a fractional Adams method.Numer. Algorithms, 36(1):31–52, 2004. 60

    work page 2004

  14. [14]

    The Volterra signature.arXiv preprint arXiv:2603.04525, 2026

    Paul P Hager, Fabian N Harang, Luca Pelizzari, and Samy Tindel. The Volterra signature.arXiv preprint arXiv:2603.04525, 2026. 1, 3, 5, 6, 7, 8, 30, 46, 47, 63

    work page arXiv 2026

  15. [15]

    Hairer, Ch Lubich, and M

    E. Hairer, Ch Lubich, and M. Schlichte. Fast numerical solution of nonlinear volterra convolution equations.SIAM journal on scientific and statistical computing, 6(3):532–541, 1985. 5

    work page 1985

  16. [16]

    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. 2

    work page 2010

  17. [17]

    Harang and Samy Tindel

    Fabian A. Harang and Samy Tindel. Volterra equations driven by rough signals.Stochastic Process. Appl., 142:34–78, 2021. 3, 7

    work page 2021

  18. [18]

    Higham.Functions of matrices

    Nicholas J. Higham.Functions of matrices. Theory and computation. Philadelphia, PA: Society for Industrial and Applied Mathematics (SIAM), 2008. 12

    work page 2008

  19. [19]

    Higham and Samuel D

    Nicholas J. Higham and Samuel D. Relton. Higher order Fréchet derivatives of matrix functions and the level-2 condition number.SIAM J. Matrix Anal. Appl., 35(3):1019–1037, 2014. 37

    work page 2014

  20. [20]

    Horn and Charles R

    Roger A. Horn and Charles R. Johnson.Matrix analysis.Cambridge: Cambridge University Press, 2nd ed. edition, 2013. 31

    work page 2013

  21. [21]

    Exponentially fading memory signature.arXiv preprint arXiv:2507.03700, 2025

    Eduardo Abi Jaber and Dimitri Sotnikov. Exponentially fading memory signature.arXiv preprint arXiv:2507.03700, 2025. 3

    work page arXiv 2025

  22. [22]

    Signatory: differentiable computations of the signature and logsig- nature transforms, on both CPU and GPU

    Patrick Kidger and Terry Lyons. Signatory: differentiable computations of the signature and logsig- nature transforms, on both CPU and GPU. InInternational Conference on Learning Representations,

    work page

  23. [23]

    Kernels for sequentially ordered data

    Franz J Király and Harald Oberhauser. Kernels for sequentially ordered data.arXiv preprint arXiv:1601.08169, 2016. 47

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  24. [24]

    Király and Harald Oberhauser

    Franz J. Király and Harald Oberhauser. Kernels for sequentially ordered data.Journal of Machine Learning Research, 20(31):1–45, 2019. 4

    work page 2019

  25. [25]

    Log-pde methods for rough signature kernels,

    Maud Lemercier, Terry Lyons, and Cristopher Salvi. Log-pde methods for rough signature kernels,

    work page

  26. [26]

    Terry J. Lyons. Differential equations driven by rough signals.Rev. Mat. Iberoam., 14(2):215–310,

    work page

  27. [27]

    Signature methods in machine learning.EMS Surv

    Andrew McLeod and Terry Lyons. Signature methods in machine learning.EMS Surv. Math. Sci., February 2025. Published online first (19 February 2025). 2

    work page 2025

  28. [28]

    Nineteen dubious ways to compute the exponential of a matrix, twenty-five years later.SIAM Rev., 45(1):3–49, 2003

    Cleve Moler and Charles Van Loan. Nineteen dubious ways to compute the exponential of a matrix, twenty-five years later.SIAM Rev., 45(1):3–49, 2003. 32, 55

    work page 2003

  29. [29]

    Igor Najfeld and Timothy F. Havel. Derivatives of the matrix exponential and their computation. Adv. Appl. Math., 16(3):321–375, 1995. 36

    work page 1995

  30. [30]

    J. M. Peña. On the multivariate Horner scheme.SIAM J. Numer. Anal., 37(4):1186–1197, 2000. 23

    work page 2000

  31. [31]

    Rubensson

    Emanuel H. Rubensson. A unifying framework for higher order derivatives of matrix functions.SIAM J. Matrix Anal. Appl., 45(1):504–528, 2024. 37

    work page 2024

  32. [32]

    The signature kernel is the solution of a Goursat PDE.SIAM Journal on Mathematics of Data Science, 3(3):873–899, 2021

    Cristopher Salvi, Thomas Cass, James Foster, Terry Lyons, and Weixin Yang. The signature kernel is the solution of a Goursat PDE.SIAM Journal on Mathematics of Data Science, 3(3):873–899, 2021. 4, 47 COMPUTATIONAL ASPECTS OF THE VOLTERRA SIGNATURE 59

    work page 2021

  33. [33]

    St. G. Samko, A. A. Kilbas, and O. I. Marichev.Fractional integrals and derivatives: theory and applications. Transl. from the Russian. New York, NY: Gordon and Breach, 1993. 27

    work page 1993

  34. [34]

    Schiff.The Laplace transform: Theory and applications

    Joel L. Schiff.The Laplace transform: Theory and applications. Undergraduate Texts Math. New York, NY: Springer, 1999. 37, 38

    work page 1999

  35. [35]

    Trefethen

    Thomas Schmelzer and Lloyd N. Trefethen. Evaluating matrix functions for exponential integrators via carathéodory–fejér approximation and contour integrals.Electronic Transactions on Numerical Analysis, 29:1–18, 2007. 39

    work page 2007

  36. [36]

    Marcel Schweitzer. Integral representations for higher-order Fréchet derivatives of matrix functions: quadraturealgorithms andnewresultsonthe level-2condition number.Linear Algebra Appl., 656:247– 276, 2023. 37, 39

    work page 2023

  37. [37]

    Scalable signature kernel computations for long time series via local neumann series expansions, 2025

    Matthew Tamayo-Rios, Alexander Schell, and Rima Alaifari. Scalable signature kernel computations for long time series via local neumann series expansions, 2025. 4

    work page 2025

  38. [38]

    Gerald Teschl.Ordinary differential equations and dynamical systems, volume 140 ofGrad. Stud. Math.Providence, RI: American Mathematical Society (AMS), 2012. 32

    work page 2012

  39. [39]

    L. N. Trefethen, J. A. C. Weideman, and T. Schmelzer. Talbot quadratures and rational approxima- tions.BIT, 46(3):653–670, 2006. 39

    work page 2006

  40. [40]

    Sulle equazioni integro-differenziali della theoria dell’elasticita.Atti Reale Accad

    Vito Volterra. Sulle equazioni integro-differenziali della theoria dell’elasticita.Atti Reale Accad. naz. Lincei. Rend. Cl. sci. fis., mat. e natur., 18:295–300, 1909. 2

    work page 1909

  41. [41]

    Gauthier-Villars, 1913

    Vito Volterra.Leçons sur les équations intégrales et les équations intégro-différentielles: Leçons pro- fessées à la Faculté des sciences de Rome en 1910. Gauthier-Villars, 1913. 2

    work page 1910

  42. [42]

    J. A. C. Weideman. Optimizing Talbot’s contours for the inversion of the Laplace transform.SIAM J. Numer. Anal., 44(6):2342–2362, 2006. 39

    work page 2006

  43. [43]

    J. A. C. Weideman and L. N. Trefethen. Parabolic and hyperbolic contours for computing the Bromwich integral.Math. Comput., 76(259):1341–1356, 2007. 39 AppendixA.Numerical V alidation The purpose of this section is to provide numerical validation of the algorithms derived in this paper, both in terms of accuracy and computational cost. We structure the di...

    work page 2007