RECORD
The Signature Kernel is the solution of a Goursat PDE
Analysis of PDEs · Submitted 26 Jun 2020
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim Here we show that for continuously differentiable paths, the signature kernel solves a hyperbolic PDE and recognize the connection with a well known class of differential equations known in the literature as Goursat problems.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
Recently, there has been an increased interest in the development of kernel methods for learning with sequential data. The signature kernel is a learning tool with potential to handle irregularly sampled, multivariate time series. In "Kernels for sequentially ordered data" the authors introduced a kernel trick for the truncated version of this kernel avoiding the exponential complexity that would have been involved in a direct computation. Here we show that for continuously differentiable paths, the signature kernel solves a hyperbolic PDE and recognize the connection with a well known class of differential equations known in the literature as Goursat problems. This Goursat PDE only depends on the increments of the input sequences, does not require the explicit computation of signatures and can be solved efficiently using state-of-the-arthyperbolic PDE numerical solvers, giving a kernel trick for the untruncated signature kernel, with the same raw complexity as the method from "Kernels for sequentially ordered data", but with the advantage that the PDE numerical scheme is well suited for GPU parallelization, which effectively reduces the complexity by a full order of magnitude in the length of the input sequences. In addition, we extend the previous analysis to the space of geometric rough paths and establish, using classical results from rough path theory, that the rough version of the signature kernel solves a rough integral equation analogous to the aforementioned Goursat PDE. Finally, we empirically demonstrate the effectiveness of our PDE kernel as a machine learning tool in various machine learning applications dealing with sequential data. We release the library sigkernel publicly available at https://github.com/crispitagorico/sigkernel.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | Recently, there has been an increased interest in the development of kernel methods for learning with sequential data. | background 1.00 | 0.04 | 0.01 | asserted 1.00 |
| s1 | The signature kernel is a learning tool with potential to handle irregularly sampled, multivariate time series. | background 0.98 | 0.10 | 0.04 | hedged possibility 0.70 |
| s2 | In "Kernels for sequentially ordered data" the authors introduced a kernel trick for the truncated version of this kernel avoiding the exponential complexity that would have been involved in a direct computation. | background 1.00 | 0.06 | 0.03 | asserted 1.00 |
| s3 | Here we show that for continuously differentiable paths, the signature kernel solves a hyperbolic PDE and recognize the connection with a well known class of differential equations known in the literature as Goursat problems. | result 0.66 | 0.95 | 0.79 | asserted 1.00 |
| s4 | This Goursat PDE only depends on the increments of the input sequences, does not require the explicit computation of signatures and can be solved efficiently using state-of-the-arthyperbolic PDE numerical solvers, giving a kernel trick for the untruncated signature kernel, with the same raw complexity as the method from "Kernels for sequentially ordered data", but with the advantage that the PDE numerical scheme is well suited for GPU parallelization, which effectively reduces the complexity by a full order of magnitude in the length of the input sequences. | contribution 0.63 | 0.92 | 0.53 | asserted 1.00 |
| s5 | In addition, we extend the previous analysis to the space of geometric rough paths and establish, using classical results from rough path theory, that the rough version of the signature kernel solves a rough integral equation analogous to the aforementioned Goursat PDE. | contribution 0.86 | 0.94 | 0.09 | asserted 1.00 |
| s6 | Finally, we empirically demonstrate the effectiveness of our PDE kernel as a machine learning tool in various machine learning applications dealing with sequential data. | result 0.62 | 0.95 | 0.10 | asserted 1.00 |
| s7 | We release the library sigkernel publicly available at https://github.com/crispitagorico/sigkernel. | contribution 0.84 | 0.83 | 0.03 | asserted 0.92 |
What was examined, and what was not
- arXiv listing abstract v9
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
Recently, there has been an increased interest in the development of kernel methods for learning with sequential data. The signature kernel is a learning tool with potential to handle irregularly sampled, multivariate time series. In "Kernels for sequentially ordered data" the authors introduced a kernel trick for the truncated version of this kernel avoiding the exponential complexity that would have been involved in a direct computation. Here we show that for continuously differentiable paths, the signature kernel solves a hyperbolic PDE and recognize the connection with a well known class of differential equations known in the literature as Goursat problems. This Goursat PDE only depends on the increments of the input sequences, does not require the explicit computation of signatures and can be solved efficiently using state-of-the-arthyperbolic PDE numerical solvers, giving a kernel trick for the untruncated signature kernel, with the same raw complexity as the method from "Kernels for sequentially ordered data", but with the advantage that the PDE numerical scheme is well suited for GPU parallelization, which effectively reduces the complexity by a full order of magnitude in the length of the input sequences. In addition, we extend the previous analysis to the space of geometric rough paths and establish, using classical results from rough path theory, that the rough version of the signature kernel solves a rough integral equation analogous to the aforementioned Goursat PDE. Finally, we empirically demonstrate the effectiveness of our PDE kernel as a machine learning tool in various machine learning applications dealing with sequential data. We release the library sigkernel publicly available at https://github.com/crispitagorico/sigkernel.
Human review
How the literature uses it
10 arXiv papers cite this in 28 sentences
Papers citing it · 21 sentences without a confident label
Uses its method or tool 2 papers
-
However, this may once again be circumvented using the signature kernel and corresponding kernel trick (see Definition [ref] and [ref] ), in the following way: use the signature kernel and kernel ridge regression [ref] to implicitly regress parameters onto the full signature, which is in a sense equivalent to using the infinitely long path signature as the candidate set of summary statistics in semi-automatic abc.
Approximate Bayesian Computation with Path Signatures · Signature Regression ABC · 0.95
-
Throughout this work, we particularly focus on the signature kernel and use the package https://github.com/crispitagorico/sigkernel sigkernel ( [ref] ), alongside PyTorch and JAX for calculating and performing functionality related to tensors and the signature kernel. https://github.com/tgcsaba/KSig KSig [ref] is another package that offers signature kernel computation.
Kernel Learning for Mean-Variance Trading Strategies · Mean-Variance Implementation · 1.00
Cites it as background 3 papers
-
We refer to [ref] for a general presentation of kernel methods with signatures and to [ref] for a kernel trick.
Framing RNN as a kernel method: A neural ODE approach · The signature · 0.90
-
Existing signature-kernel implementations [ref] compute geometric signatures, whose semimartingale limits correspond to iterated Stratonovich integrals.
Learning Manifold and It\^o Dynamics with Branched Neural Rough Differential Equations · The Branched Signature Kernel · 0.80
-
A more recent method to compute untruncated signature kernels using PDEs was developed by Cass et al. [ref] .
Signatures, Lipschitz-free spaces, and paths of persistence diagrams · Previous and Related Work · 0.87
What it cites · 9 sentences without a confident label
Uses the method or tool of 1 paper
-
Following [ref] , we consider the problem of reducing the size of the support in $𝒳$ of a discrete measure $μ$ whilst preserving some of its statistical properties.
A Randomized Algorithm to Reduce the Support of Discrete Measures · Moments-matching reduction algorithm for the support of a discrete measure on paths · 0.81
Cites as background 2 papers
-
Finally, we note that the article [ref] first treated the truncated signature kernel in the case of branched rough paths.
Signature moments to characterize laws of stochastic processes · Introduction · 0.83
-
Integration of two parameters rough integrals is also discussed in [ref] .
Rough sheets · Introduction · 0.96
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.