pyunlocbox 0.2.0

A convex optimization toolbox using proximal splitting methods.

PyUNLocBoX is a convex optimization toolbox using proximal splitting methods implemented in Python. It is a free software distributed under the BSD license and is a port of the Matlab UNLocBoX toolbox.

• Code : https://github.com/epfl-lts2/pyunlocbox

• Documentation : http://pyunlocbox.readthedocs.org

• PyPI package : https://pypi.python.org/pypi/pyunlocbox

• Travis continuous integration : https://travis-ci.org/epfl-lts2/pyunlocbox

• UNLocBoX matlab toolbox : http://unlocbox.sourceforge.net

Features

• Solvers

  • Forward-backward splitting algorithm

  • Douglas-Rachford splitting algorithm

• Proximal operators

  • L1-norm

  • L2-norm

  • Projection on the L2-ball

Installation

PyUnLocBox is continuously tested with Python 2.6, 2.7, 3.2, 3.3 and 3.4.

System-wide installation:

$ pip install pyunlocbox

Installation in an isolated virtual environment:

$ mkvirtualenv --system-site-packages pyunlocbox
$ pip install pyunlocbox

You need virtualenvwrapper to run this command. The --system-site-packages option could be useful if you want to use a shared system installation of numpy and matplotlib. Their building and installation require quite some dependencies.

Another way is to manually download from PyPI, unpack the package and install with:

$ python setup.py install

Execute the project test suite once to make sure you have a working install:

$ python setup.py test

Authors

PyUNLocBoX was started in 2014 as an academic project for research purpose at the LTS2 laboratory from EPFL. See our website at http://lts2www.epfl.ch.

Development lead :

• Michaël Defferrard from EPFL LTS2 <[e-mail]>

• Nathanaël Perraudin from EPFL LTS2 <[e-mail]>

Contributors :

• None yet. Why not be the first ?