RoughPy: streaming data is rarely smooth

Conference item

Rough path theory is a branch of mathematics arising out of stochastic analysis. One of the main tools of rough path analysis is the signature, which captures the evolution of an unparametrised path including the order in which events occur. This turns out to be a useful tool in data science applications involving sequential data. RoughPy is our new Python package that aims change the way we think about sequential streamed data, by viewing it through the lens of rough paths. In RoughPy, data is wrapped in a stream object which can be composed and queried to obtain signatures that can be used in analysis. It also provides a platform for further exploration of the connections between rough path theory and data science.

Authors: Morley, S; Lyons, T

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: NumFOCUS - Insight Software Consortium (ITK)
• Host title: Proceedings of the 23rd Python in Science Conference
• Pages: 320-331
• Series: Proceedings of the Python in Science Conferences
• Publication date: 2024-07-10
• Event title: 23rd Python in Science Conference (SciPy 2024)
• Event location: Tacoma, WA, USA
• Event website: https://www.scipy2024.scipy.org/
• Event start date: 2024-07-08
• Event end date: 2024-07-14
• DOI: 10.25080/dxwy3560
• EISSN: 2575-9752
• Language: English
• Keywords: time series; unparametrised paths; data science; machine learning; signatures; signature kernels; sequential data; Log-ODE method; rough paths