Installation

Julia

Julia 1.10 or later is required. General registration is pending. Install the self-contained Julia subpackage from GitHub:

using Pkg
Pkg.add(url="https://github.com/Xunjian-Li/SpectralMM", subdir="julia")
Pkg.add(["Distributions", "GLM", "StatsModels"])

Load it with using SpectralMM. Pure Julia is the default backend. An explicit backend=:cpp call builds and caches the native library, requiring a C++17 compiler. CMake and Eigen artifacts are currently unconditional dependencies; installing the package downloads them even if C++ is never used.

Python

python -m pip install SpectralMM

import numpy as np
import spectralmm
rng = np.random.default_rng(71)
X = rng.normal(size=(100, 3))
y = rng.binomial(1, 0.5, size=100)
model = spectralmm.glm(X, y, family="binomial", link="logit")
print(model.summary())

The source distribution requires a C++17 compiler. pip installs build tools and runtime dependencies automatically. Formula input uses Patsy.

R

install.packages("remotes")
remotes::install_github("Xunjian-Li/SpectralMM", subdir="R")
library(SpectralMM)
set.seed(71)
X <- matrix(rnorm(300), 100, 3)
y <- rbinom(100, 1, 0.5)
model <- spectralmm_glm(X, y, family=binomial("logit"))
summary(model)

Source installation uses R's build system and requires a C++17 compiler (Rtools on Windows). Rcpp and RcppEigen are package dependencies. CRAN registration is pending. Use help(package="SpectralMM") for installed help.

Local development

Run these commands from the repository root to install a language package:

python -m pip install ./python
R CMD INSTALL R
julia --project=julia -e 'using Pkg; Pkg.instantiate()'

R source installation requires Rcpp and RcppEigen in the active R library. Use julia --project=julia to start a Julia session in the local package environment. To use the local Julia package from another environment (such as a notebook), run Pkg.develop(path="/absolute/path/to/SpectralMM/julia") in that environment.

Language code is maintained in R/, python/ and julia/. Shared C++ files are maintained in root cpp/; synchronize their bundled copies with python3 tools/sync_package_sources.py before committing.