Fast, parallelized implementation of Principal Component Analysis with constant memory consumption for large data sets.

View the Project on GitHub lettis/FastPCA


FastPCA is a PCA-calculator programmed in C++(11). Computation is parallelized with OpenMP.

For fast matrix diagonalization, LAPACK is used (and needed, of course).

The project includes the 'xdrfile' library of GROMACS. Thus, you can use data files written as ASCII data as well as .xtc-trajectories.

For bug-reports, write to or

Happy Computing.


The code is published "AS IS" under the simplified BSD license. For details, please see LICENSE.txt

If you use the code for published works, please cite as

"Florian Sittel (2016) FastPCA - fast Principal Component Analysis Package,"


Create a build-directory in the project root and change into that directory: # mkdir build # cd build Run cmake, based on the underlying project: # cmake .. Hopefully, everything went right. If not, carefully read the error messages. Typical errors are missing dependencies...

If everything is o.k., run make (on multicore machines, use '-j' to parallelize compilation, e.g. 'make -j 4' for up to four parallel jobs): # make Now, you should find the 'fastca' binary in the 'src' folder.