Supervised learning of analysis-sparsity priors with automatic differentiation

This repo contains the python scripts that reproduce the figures in the published paper.

As we also provide the package we built to learn analysis-sparsity priors, solving a bilevel optimization with Automatic Differentiation, in the directory <./modules>, these scripts serve as a guiding example to help users applying our framework on other datasets.

Dependencies

First install needed packages using Conda by running:

conda env create -f environment.yml

Then, activate the created environment called tv:

conda activate tv

Running a script:

To run the script that generates a figure in the published paper, execute the following in the command line:

python -u name_of_script.py

after replacing name_of_script by the name of the according script from the three ones we provide here. To reproduce figure 1 for example :

python -u plot_fig1_varying_noise_amplitude.py