plskit (R)

R package for plskit: Partial Least Squares with modern inference, backed by a Rust engine.

What is plskit?

plskit is a Partial Least Squares (PLS) library built around four ideas:

• Modern inference. Confirmatory tests use split-half held-out prediction: split_exact is calibrated by permutation, and split_nb is its faster Fisher-z approximation. Confidence intervals are taken on rotation-invariant quantities (held-out correlation, per-variable leverage, a subsampling z for β), so no Procrustes alignment is needed. The tests are introduced in our NeurIPS 2026 paper.
• Same results in every language. All numerical work runs in one Rust engine. The Python and R wrappers call into it, and the Julia wrapper runs the Python package, so a fixed (version, seed, X, y) gives the same results, within a bit-near tolerance, in every language and on every supported platform.
• Speed. The engine is written in Rust and fits PLS1 with Improved Kernel PLS (Dayal and MacGregor 1997), which makes plskit probably the fastest PLS package on a single core.
• Parallel and RAM-safe. Inference runs efficiently in parallel without blowing up RAM.

Install

Requires R 4.2 or newer. Install a release from r-universe:

install.packages("plskit", repos = "https://pawlenartowicz.r-universe.dev")

or from the r-universe branch of the GitHub repository:

devtools::install_github("pawlenartowicz/plskit", ref = "r-universe")

r-universe serves binaries for Windows, macOS and Linux, which need no Rust. A source install (from GitHub, or from r-universe where it has no binary for your R version or platform) needs Rust 1.85 or newer with cargo, from rustup; on Windows it also needs Rtools and the GNU Rust target (rustup target add x86_64-pc-windows-gnu).

A checkout of main builds only in dev mode: a cargo config file outside the package points the engine crates at the checkout.

git clone https://github.com/pawlenartowicz/plskit
cd plskit
printf "[patch.crates-io]\nplskit = { path = '%s/plskit-rs' }\nplskit-bind = { path = '%s/plskit-bind' }\n" "$PWD" "$PWD" > "$HOME/plskit-dev.toml"
PLSKIT_CARGO_CONFIG="$HOME/plskit-dev.toml" R CMD INSTALL plskit-r

See installation for what the dev-mode build does and how to fix a failed build.

Quickstart

library(plskit)

set.seed(1)
X <- matrix(rnorm(60 * 6), 60, 6)
y <- 4 * (X[, 1] + X[, 2]) + 0.5 * rnorm(60)

# Fit and predict
fit  <- pls1_fit(X, y, k = 3, seed = 42)
yhat <- pls1_predict(fit, X[1:5, ])

# Confirmatory test for any signal; k = 1 with split_exact is recommended.
test <- pls1_confirmatory_test(X, y, k = 1, test_method = "split_exact",
                               seed = 42)
test$pvalue

Same names as Python

Function names, argument names and result fields are the Python ones. The number of components is k (and k_max), as in every plskit language, not ncomp as in the R pls package; there is no alias.

R value types

• Results are named lists with an S3 class: pls1_fit returns a c("pls1_result", "plskit_result"). A 1-D ndarray becomes a numeric vector, a 2-D one a numeric matrix, and None a retained NULL element.
• Seeds are decimal strings, because R has no unsigned 64-bit integer: result$seed is a character scalar, and seed = result$seed reproduces the run exactly.
• Errors are R conditions carrying the Python error codes: plskit_error, with the subclasses plskit_invalid_weights and plskit_resampling_degenerate.

tryCatch(pls1_fit(X, y, k = 2.5), plskit_error = function(e) e$code)
# "invalid_argument"

The full mapping is in Differences from Python.

Documentation

• R pages: installation, quickstart, differences from Python.
• Python API and result objects: the canonical reference for every function, argument and result field.

License

GPL-3.0-or-later.

Citation

@inproceedings{lenartowicz2026cheap,
  title     = {Cheap and Powerful Tests for Supervised Subspaces: Per-Component Inference for {PLS}},
  author    = {Lenartowicz, Pawe{\l} and Plisiecki, Hubert},
  booktitle = {The Fortieth Annual Conference on Neural Information Processing Systems},
  year      = {2026},
  url       = {https://openreview.net/forum?id=xb6CB7d9LO}
}

---

Paweł Lenartowicz — Freestyler Scientist · GitHub · ORCID