FastPCA: An R package for fast singular value decomposition.

Creator(s)

• Kimberly R. Ward
• Mitchell Hayes
• Lauren C. Peres
• Brooke L. Fridley, Children's Mercy Kansas CityFollow
• Steven Eschrich
• Alex C. Soupir

Document Type

Article

Publication Date

3-7-2026

Identifier

DOI: 10.21105/joss.09706; PMCID: PMC13048341

Abstract

The FastPCA package provides an interface to optimized matrix multiplication libraries (libtorch) for the purpose of singular value decomposition. Using FastPCA to perform randomized singular value decomposition (SVD, (Halko, Martinsson, & Tropp, 2011)) with the torch or pytorch backend drastically reduces the computational time compared to base R prcomp and other truncated singular value decomposition. Developed for biological data such as single-cell RNA-sequencing, spatial transcriptomics, or matrix-assisted laser desorption/ionization (MALDI imaging), FastPCA can efficiently and accurately identify the leading singular values in high-dimensional data.

Journal Title

J Open Source Softw

Volume

11

Issue

119

PubMed ID

41937916

Comments

Grants and funding

• R01 CA279065/CA/NCI NIH HHS/United States

This work is licensed under a Creative Commons Attribution 4.0 International License.

Publisher's Link: https://joss.theoj.org/papers/10.21105/joss.09706

Recommended Citation

Ward KR, Hayes M, Peres LC, Fridley BL, Eschrich S, Soupir AC. FastPCA: An R package for fast singular value decomposition. J Open Source Softw. 2026;11(119):9706. doi:10.21105/joss.09706