/pcp

A Python implementation of the Principal Component Pursuit algorithm from arXiv:0912.3599

Primary LanguagePythonMIT LicenseMIT

Principal Component Pursuit in Python

This is a Python implementation of the Principal Component Pursuit algorithm for robust PCA.

This implementation uses the fbpca implementation of approximate partial SVD for speed so you'll need to install that first.

Usage

TODO

Demo

Applied to the 'Escalator' dataset (using the code in the demo.py script, this algorithm produces a video with frames that look like:

https://raw.githubusercontent.com/dfm/pcp/master/demo.png

Author & License

Copyright 2015 Daniel Foreman-Mackey

This is open source software written by Dan Foreman-Mackey and released under the terms of the MIT license (see LICENSE).