fast.EOT

Fast version (parallelized) to calculate EOTs (Empirical orthogonal functions) in R. It is based on the 'eof' function from remote, which already utilises efficient C++ code, but implements an additional parallelization step, making the calculation of EOTs approximately 7x faster. Currently 'fast.EOT' only calculates the EOTs of a single spatio-temporal field, contrary to the 'eot' function in 'remote', which also allows to calculate EOTs based on two seperate fields (predictor & predictant).

To load (using devtools):

library(devtools)
install_github("BernhardKuehn/fast.EOT")

to cite:

citation("fast.EOT")

Non-CRAN required packages

Utilising C++ code for speed in a parallelized way, requires the installation of an additional R package (Rcpp2doParallel) that is not (yet) available on CRAN. The Rcpp2doParallel R package allows the implementation of C++ functions and parallelization using the doParallel and foreach backend. Therefore the overhead of compiling C++ code on each parallel worker is reduced, allowing a faster code execution.

# Non-CRAN
# an Rcpp-extension that handles C++ code in a parallel backend
devtools::install_github("coatless-rd-rcpp/rcpp-and-doparallel")
library("Rcpp2doParallel")