PPCA

Probabilistic PCA (PPCA) + MPPCA written in torch.

Overview

ppca-torch implements in PyTorch the method Probabilistic Principal Component Analysis (PPCA) as described by Tipping & Bishop (1999). It provides also a PCA implementation and the sklearn implementation ("baseline"). Additionaly, it contains also the extension to MPPCA. Examples and notebooks allow to see:

• comparison between different ways to learn the parameters of the model (convergence and projections)
• sampling examples from MNIST
• comparison of projections on PCA vs. PPCA
• missing values impact with MPPCA

The repository is organised as follows:

• src/ppca contains the source code for PPCA, MPPCA and PCA implementations.
• notebooks contains visual experiments
• examples show more scripts that enable to test various things like compare the convergence of methods, sample from a model trained on MNIST, compare PCA and PPCA etc.
• figures have some interesting plots

Quickstart

Dev mode

First, create (python3 -m venv .venv) or activate a virtualenv source .venv/bin/activate.

Then clone the repository and install:

git clone https://github.com/siemovit/ppca.git
cd ppca
pip install -e .

References

Tipping, Michael E., and Christopher M. Bishop. "Probabilistic principal component analysis." Journal of the Royal Statistical Society Series B: Statistical Methodology 61.3 (1999): 611-622.