FastRandPCA.jl

Coverage

Fast randomized PCA algorithms for Sparse Data. It's a julia re-implementation of Matlab frPCA_sparse.

The package includes an implementation of two functions eig_svd and pca, which are designed to work with sparse matrices (SparseMatrixCSC), though can also be applied with normal ones (Matrix).

Example code:

using SparseArrays
eig_svd(sprand(100, 100, 0.2), 10)

Installation

To install it, run ] add FastRandPCA or import Pkg; Pkg.add("FastRandPCA") from Julia REPL. Use ] add FastRandPCA#master if you want to install a development version.

References

  • Xu Feng, Yuyang Xie, Mingye Song, Wenjian Yu, and Jie Tang, "Fast Randomized PCA for Sparse Data," in Proc. the 10th Asian Conference on Machine Learning (ACML), Beijing, China, Nov. 2018. 10.48550/arXiv.1810.06825.