pysersic is a code for fitting sersic profiles to galaxy images using Bayesian Inference. It is written in python using jax with inference performed using numpyro. You can read more about the motivation for this software via the JOSS Paper.

Installation

pysersic now requies at least python version 3.9. To start you should have jax installed. The easiest way to do this is to use pip install 'jax[cpu]', however if you are on windows or planning on running it on a GPU, the installation can be a little more complicated, please see the guide here. The other package dependencies for pysersic are: - arviz - asdf - astropy - corner - matplotlib - numpy - numpyro - pandas - photutils - scipy - equinox - interpax - tqdm

Optional - jaxopt (only when using the pure minimization option).

There is a provided minimal environment.yml capturing the above in the base repo, as well as an environment-lock.yml which has a full set of versions specified for every dependency --- if you are having trouble with the installation, creating an environment using this file should ensure you have a single set of package versions that pysersic is known to work with. (We don't recommend using the lock yml in general, as pysersic should be kept up to date with new versions of its dependencies.)

Next you can install pysersic! To install the latest stable release, you can simply pip install via

pip install pysersic If you would like the bleeding edge, development build (or would like to modify, add to, or otherwise access pysersic source files), you can also install from the github:

git clone https://github.com/pysersic/pysersic.git cd pysersic pip install -e .

Basic usage

All you need to run pysersic is a cutout of you galaxy of interest, and cutout of the error map, pixelized version of the PSF, and optionally a mask specifying bad pixels or nearby sources. The most basic setup is shown below for many many more details and in depth examples please see the documentation. First we set up a prior.

``` from pysersic.priors import SourceProperties

props = SourceProperties(imgcutout,mask=mask) # Optional mask prior = props.generateprior('sersic', # Other profiles inclues 'exp', 'dev' and 'pointsource' sky_type='none') # Can also use 'flat' or 'tilted-plane' to simultaneously fit a background ``` Then we initialize the fitting object:

``` from pysersic import FitSingle from pysersic.loss import gaussian_loss

fitter = FitSingle(data=imgcutout, # Cutout of galaxy rms=sigcutout, # Cutout of pixel errors psf=psf, #Point spread function prior=prior, # Prior that we initialized above mask=mask, #Optional mask lossfunc=gaussianloss) # Can specify loss function! See loss.py for many options or write your own! ```

Now we can fit!

``` mapdict = fitter.findMAP() # Find the 'best-fit' parameters as the maximum-a-posteriori. Returns a dictionary containing the MAP parameters and model image

svires = fitter.estimateposterior('svi-flow') #Train a normalizing flow to estimate the posterior, retruns a PysersicResults object containing the posterior and other helper methods nuts_res = fitter.sample() #Sample the posterior using the No U-turn sampler (NUTS), retruns a PysersicResults object ```

Test Suite

There is a growing set of tests that can be used to confirm the behavior of pysersic and its features is as expected. These tests run automatically on each pull request and merge. If you would like to run the tests yourself on your local machine, you can do so with the pytest package. Once installed, you can simply navigate to the tests directory and run:

pytest

Citation

If you use pysersic and find it useful, please do cite it. You can use the following BibTex.

@article{Pasha2023, doi = {10.21105/joss.05703}, url = {https://doi.org/10.21105/joss.05703}, year = {2023}, publisher = {The Open Journal}, volume = {8}, number = {89}, pages = {5703}, author = {Imad Pasha and Tim B. Miller}, title = {pysersic: A Python package for determining galaxy structural properties via Bayesian inference, accelerated with jax}, journal = {Journal of Open Source Software} }