VBPCApy

Variational Bayesian PCA with missing data support.

VBPCApy implements Variational Bayesian Principal Component Analysis (Ilin & Raiko, 2010) with native support for incomplete observations, sparse masks, and posterior uncertainty quantification. It provides a scikit-learn-compatible estimator, missing-aware preprocessing utilities, and empirical model selection for the number of latent components.

Quick install

pip install vbpca-py

Minimal example

import numpy as np
from vbpca_py import VBPCA

x = np.random.randn(50, 200)  # 50 features × 200 samples
model = VBPCA(n_components=5, maxiters=100)
scores = model.fit_transform(x)
recon = model.inverse_transform()

Next steps

• Installation — full install guide with extras and build-from-source instructions.
• Quick Start — dense and sparse examples with annotations.
• Tutorials — narrative walkthroughs for common workflows.
• API Reference — complete public API documentation.