Hierarchical FISTA - phase retrieval for pulsar spectroscopy
For a full description please see the published paper.
The main dependencies of our code include the widely used libraries such as NumPy, SciPy, and Astropy. For plotting, we use Matplotlib. Thus, it is possible you will be able to run H-FISTA without the need to install any extra libraries.
However, if you want do need to install dependencies, or wish to ensure the exact same versions of packages are used, we provide an easy way to do so. We recommend using python poetry for installing the dependencies:
poetry install
Alternatively, you can use pip:
pip -r requirements.txt
You might need to install additional LaTeX fonts used by matplotlib; e.g.
sudo apt install cm-super
And if using jupyter notebooks, you might also need to install dvipng; e.g.
sudo apt install dvipng
The data used in the publication and in the example notebooks can be obtained from Zenodo under this DOI: 10.5281/zenodo.7007226
H-FISTA accepts input data in ASCII format (as produced by psrflux, part of PSRchive), a FITS file, or a pickle. If using a notebook, you can use any custom reader as long as the data ends up in a NumPy array.
To see how we used the code to produce the results in the paper, please see the notebooks included in this repostitory:
- Simulated noise-free case with 12.5 per cent local density
- Simulated noise-free case with 25.0 per cent local density
- Simulated noisy case with 12.5 per cent local density
- Observation of PSR J0837+0610
- Observation of PSR J1939+2134
It is also possible to use the code on the command line. For example to reproduce the results for J0837+0610:
python HFISTA.py --data J0837+0610.fits --striation
Contributions are welcome. Please use black for linting and, if possible, include tests.