Signatory: differentiable computations of the signature and logsignature transforms, on both CPU and GPU

Signatory is a library for calculating signature and logsignature transforms and related functionality. The focus is on making this functionality available for use in machine learning, and as such includes features such as GPU support and backpropagation. To our knowledge it is the first publically available GPU-capable library for these operations. It also implements several new algorithmic improvements, and provides several new features not available in previous libraries. The library operates as a Python wrapper around C++, and is compatible with the PyTorch ecosystem. It may be installed directly via \texttt{pip}. Source code, documentation, examples, benchmarks and tests may be found at \texttt{\url{this https URL}}. The license is Apache-2.0.

[1]  Lianwen Jin,et al.  Rotation-free online handwritten character recognition using dyadic path signature features, hanging normalization, and deep neural network , 2016, 2016 23rd International Conference on Pattern Recognition (ICPR).

[2]  Terry Lyons,et al.  Learning stochastic differential equations using RNN with log signature features , 2019, ArXiv.

[3]  M. Lothaire.  Combinatorics on words: Bibliography , 1997 .

[4]  Lianwen Jin,et al.  Chinese character-level writer identification using path signature feature, DropStroke and deep CNN , 2015, 2015 13th International Conference on Document Analysis and Recognition (ICDAR).

[5]  James Morrill,et al.  The Signature-Based Model for Early Detection of Sepsis From Electronic Health Records in the Intensive Care Unit , 2019, 2019 Computing in Cardiology (CinC).

[6]  Benjamin Graham,et al.  The iisignature library: efficient calculation of iterated-integral signatures and log signatures , 2017, ACM Trans. Math. Softw..

[7]  Jeremy Francis Reizenstein.  Iterated-integral signatures in machine learning , 2019 .

[8]  Franz J. Király,et al.  Kernels for sequentially ordered data , 2016, J. Mach. Learn. Res..

[9]  Lianwen Jin,et al.  DeepWriterID: An End-to-End Online Text-Independent Writer Identification System , 2015, IEEE Intelligent Systems.

[10]  Terry Lyons.  Rough paths, Signatures and the modelling of functions on streams , 2014, 1405.4537.

[11]  Arun Ram,et al.  Standard Lyndon bases of Lie algebras and enveloping algebras , 1995 .

[12]  Andrey Kormilitzin,et al.  A Primer on the Signature Method in Machine Learning , 2016, ArXiv.

[13]  Xin Zhang,et al.  LPSNet: A Novel Log Path Signature Feature Based Hand Gesture Recognition Framework , 2017, 2017 IEEE International Conference on Computer Vision Workshops (ICCVW).

[14]  Terry Lyons.  Di erential equations driven by rough signals , 1998 .

[15]  C. Reutenauer.  Free Lie Algebras , 1993 .