Q-MM’s documentation

Q-MM is a small Python toolbox to optimize differentiable objective functions

\[\hat x = \underset{x \in \mathbb{R}}{\text{arg min}}\ J(x)\]

by Majorization-Minimization with quadratic surrogate function. In particular, no linesearch is necessary and close form formula for the step are used with guaranteed convergence without sub-iteration. The explicit step formula allows fast convergence of the algorithm to a minimizer with minimal tuning parameters. However, the objective function must meet the conditions, see Background.


  • The mmmg (or 3mg), Majorize-Minimize Memory Gradient algorithm.

  • The mmcg, Majorize-Minimize Conjugate Gradient algorithm.

  • No linesearch: the step is obtained from a close form formula.

  • No conjugacy choice: a conjugacy strategy is not necessary thanks to the subspace nature of the algorithms. The mmcg algorithm use a Polak-Ribière formula.

  • Generic and flexible: there is no restriction on the number of regularizer, their type, .., as well as for data adequacy.

  • Provided base class for objective allowing easier and fast implementation.

  • Just one file if you like quick and dirty installation, but available with pip.

  • Comes with examples of implemented linear operator.


The code is hosted on Github under GPLv3 License. Feel free to contribute or submit issue.


If you are having issues, please let us know

orieux AT l2s.centralesupelec.fr

F. Orieux (academic pages) and R. Abirizk are affiliated to the Signal and Systems Laboratory L2S.


Authors would like to thanks Jérôme Idier, Saïd Moussaoui and Émilie Chouzenoux. E. Chouzenoux has also a Matlab package that implements 3MG for image deconvolution on here webpage.


The project is licensed under the GPLv3 license. If you use the library, please cite it

   title = {Q-MM: The Quadratic Majorize-Minimize Python toolbox},
   author = {Orieux, Fran\c{c}ois and Abirizk, Ralph},
   url = {https://github.com/forieux/qmm},

and associated publications, see Background.