regpy: Python tools for regularization methods

For information and detailed documentation, please visit https://num.math.uni-goettingen.de/regpy/.

A library for implementing and solving ill-posed inverse problems developed at the Institute for Numerical and Applied Mathematics Goettingen.

This project is now low beta quality software and still under under heavy development. Therefore, expect bugs and partially undocumented tools.

Usage examples

To get an impression of how using regpy looks, there are some examples in the examples folder on GitHub, as well as inside the release tarballs (see below). Most of the examples supply both a commented python script and a python notebook with more detailed explanation.

Installation

Obtaining the source code

• The source code is on GitHub (regpy/regpy).
• Releases are at the corresponding release page. The current version is 0.2.

Dependencies

• numpy >= 1.14
• scipy >= 1.1

Optional dependencies

• ngsolve (for some forward operators that require solving PDEs)
• bart (for the MRI operator)
• matplotlib (for some of the examples)
• pdoc3 (for generating the documentation)

Installation with pip

A basic setup.py is provided, but the package is not on PyPI yet. To install it, clone this repository or download the release tarball, and run

pip install .

from the project's root folder. If you want to modify regpy itself, you can use

pip install --editable .

to have Python load regpy from your current directory rather than copying it to its library folder.

In both cases, you can add --no-deps to prevent installing dependencies automatically, in case you want to install them via some other package manager.