Optimized Parallelized Matrix Pearsons Correlation Coefficient
The code presented here is an attempt to provide an algorithm to perform Pearsons Correlation Coefficient calculations at a large scale for data sets arranged as rows or columns in rectangular matrices. This particular algorithm was designed to be performant in the presence of missing data.
The initial code considers implementation on a multicore (single node) shared memory machine with thread level parallelism and vectorization capability. In future versions of this code, we hope to expand the algorithm to include distributed memory parallelism.
Initial work has been done to compute using GPUs, more information can be found HERE
MPCC is written in C and as such can be used from any programming language which supports calling functions via the Foreign Function Interface (FFI). Bindings for the R language for statistical computing are available.
MPCC can be used directly from the R language for statistical computing. To learn how to install it please visit the R instructions.
MPCC can be directly used in C/C++ projects, and other languages via the Foreign Function Interface (FFI) please visit the C/C++ instructions for more information.
MPCC comes with an extensive test-suite to make sure the results are identical to correlation coefficients computed by e.g. the cor() function in R. The R regression test-suite can be found (here)[tests/]
Issues can be raised through the github issue tracker.
Want to contribute? Great! Contribute to the MPCC source code by forking the Github repository, and sending us pull requests. Its also possible to just post comments on code / commits. Or be a maintainer, and adopt a function
Questions can be send via email to: Danny Arends Danny.Arends@gmail.com, Chad Burdyshaw chadburdyshaw@gmail.com, Glenn Brook glenn-brook@tennessee.edu