Towards a topological pattern detection in fluid and climate simulation data



Muszynski, Grzegorz, Kashinath, Karthik, Kurlin, V ORCID: 0000-0001-5328-5351, Wehner, Michael and Prabhat,
(2018) Towards a topological pattern detection in fluid and climate simulation data. In: Climate Informatics, 2018-9-19 - 2018-9-21, Boulder, Colorado, US.

[img] Text
pattern_detection_climate.pdf - Author Accepted Manuscript
Access to this file is restricted: awaiting official publication and publisher embargo.

Download (4MB)

Abstract

Increasingly massive amounts of high- resolution climate datasets are being generated by observations as well as complex climate models. As the unprecedented growth of data continues, a massive challenge is to design automated and efficient data analysis techniques that can extract meaningful insights from vast datasets. In particular, a key challenge is the detection and characterization of weather and climate patterns. Machine learning, including deep learning, are currently popularly used for these tasks. These techniques, however, do not incorporate geometric features of data and temporal persistence information. In this paper, we develop a novel approach to pattern detection and characterization based on dynamical systems, manifold learning and topological data analysis (i.e., persistent homology) that utilize important geometric and topological properties of underlying patterns in datasets.

Item Type: Conference or Workshop Item (Unspecified)
Depositing User: Symplectic Admin
Date Deposited: 25 Oct 2018 08:56
Last Modified: 19 Jan 2023 01:15
Related URLs:
URI: https://livrepository.liverpool.ac.uk/id/eprint/3026927