A Python library for Probabilistic PCA that handles missing data without asking you to impute first. It uses the EM algorithm under the hood, with the number-crunching done in Rust.
Regular PCA doesn't cope well when your data has holes in it. Probabilistic PCA treats missing values as latent variables in a generative model — a more principled approach than patching NaNs and hoping for the best.
For the full story, see Tipping & Bishop (1999), "Probabilistic Principal Component Analysis", Journal of the Royal Statistical Society, 61(3), 611–622.
- Rust backend — keeps things snappy on larger datasets
- Missing value support — the main reason this exists
- Scikit-learn compatible — fits in wherever you'd use
sklearn.decomposition.PCA
pip install rustypcaYou'll need a Rust toolchain to build from source. Python >= 3.10.
import numpy as np
from rustypca import PPCA
X = np.random.randn(100, 10)
X[np.random.rand(100, 10) < 0.1] = np.nan # Introduce some missing values
model = PPCA(n_components=2)
X_transformed = model.fit_transform(X)
X_reconstructed = model.inverse_transform(X_transformed)No preprocessing or imputation needed.
make testThis project was built with the help of Claude.
MIT
