PNPD (Preconditioned Nested Primal-Dual)

This repository implements the Preconditioned Nested Primal-Dual (PNPD) method. It includes both the core PNPD package and some examples for testing its performance.

Repository Structure

├── examples
│   ├── example_1
│   ├── example_2
│   └── example_3
└── PNPD
    ├── math_extras.py
    ├── plot_extras.py
    ├── schedulers.py
    ├── solvers.py
    └── utilities.py

• PNPD: python package and supporting utilities.
  • solvers.py: implementation of the PNPD method and other solvers used for the examples.
  • math_extras.py: mathematical functions and utilities.
  • plot_extras.py: utilities for generating plots of results.
  • schedulers.py: scheduling functions for the non-stationary version of PNPD.
  • utilities.py: other generic utilities.
• examples: test scripts and examples for evaluating the PNPD method's performance.
  • example_*: each folder contains a suite of tests for a specific problem.

Running Examples

1. (Optional, recommended) Create a virtual environment to isolate dependencies (see python docs)

2. Move to the root folder of this repository:

   cd path/to/repo
   

3. Install requirements with

   pip install .
   

4. Move to an example folder (e.g., example_1):

   cd ./examples/example_1
   

5. Generate the blurred image and point spread function (PSF) for the example

   python init.py
   

6. Run a test, (e.g., PNPD_comparison.py):

   python PNPD_comparison.py 
   

7. Check the generated plots and results in ./examples/plots/ directory.