Summary

the 2014 International Symposium on Nonlinear Theory and its Applications

2014

Session Number:B2L-B

Session:

Number:B2L-B5

Fast Recurrence Quantification Analysis on GPUs

Tobias Rawald,  Mike Sips,  Norbert Marwan,  Doris Dransch,  

pp.325-328

Publication Date:2014/9/14

Online ISSN:2188-5079

DOI:10.34385/proc.46.B2L-B5

PDF download (343KB)

Summary:
In this paper we present a novel computing approach for conducting recurrence quantification analysis. It utilises the concepts of Divide and Recombine to distribute the detection of vertical and diagonal lines in a recurrence matrix among multiple graphics card processors. We employ our approach to calculate recurrence quantification measures based on an hourly air temperature record of Potsdam, a time series with more than one million data points. The proposed calculation scheme reduces the computing time drastically; from more than 6 h using a singlethreaded CPU implementation to about 5 min using two graphics card processors.