pacpack

pcapack is a high-performance C library with R bindings that can be used to quickly compute principal components, including truncated methods.

Note that the package is currently under development, and is not particularly stable.

Benchmarks

Covariance:

     test replications elapsed relative
2 pcapack           10   0.182    1.000
1       R           10   3.444   18.923

PCA:

     test replications elapsed relative
1 pcapack           10   2.267    1.000
2       R           10   3.350    1.478

SVD:

     test replications elapsed relative
1 pcapack           10   1.422    1.000
2       R           10   1.725    1.213

Centering and Scaling:

### center=TRUE, scale=FALSE
     test replications elapsed relative
2 pcapack           10   0.236    1.000
1       R           10   2.381   10.089

### center=FALSE, scale=TRUE
     test replications elapsed relative
2 pcapack           10   0.277    1.000
1       R           10   5.207   18.798

### center=TRUE, scale=TRUE
     test replications elapsed relative
2 pcapack           10   0.476    1.000
1       R           10   7.719   16.216

You can find the source for these benchmarks in the inst/benchmarks/ tree. All tests performed using:

• R 3.2.0
• OpenBLAS
• gcc 4.9.1
• 4 cores of a Core i5-2500K CPU @ 3.30GHz

Requirements and Installation

To install, you will need:

• cmake >= 2.8.1
• A C99 compatible compiler with OpenMP >= 4 support.
• LAPACK and BLAS libraries (will use R's if installing the R package)
• R >= 2.14.0 and the RNACI package (if installing the R package)

Both the R package and the standalone library require cmake.

To install the R package, simply execute:

library(devtools)
install_github("wrahtematics/RNACI") ### dependency
install_github("wrathematics/pcapack")

To build just the shared library, in your terminal, execute make in pcapack/src/pcapack/. A static and dynamic library will be placed in the pcapack/src/pcapack/build tree.