The JAX-accelerated Universal Bayesian Imaging Kit is a python package for high-fidelity Bayesian imaging.

J-UBIK allows to image observations from different instruments with Bayesian posterior uncertainties and component separation. Next to many useful generic tools and building blocks, JUBIK comes with a series of sky models and instrument implementations, namely:

• Chandra
• eROSITA
• James Webb Space Telescope

This package can be installed via pip.

git clone https://github.com/NIFTy-PPL/J-UBIK
cd j-ubik
pip install --user .

for a regular installation. For editable installation add the -e flag.

• JAX
• astropy
• NIFTy (follow installation for NIFTy.re, the JAX implementation of NIFTy)
• ducc0
• matplotlib

For testing you need pytest to be installed. To run the tests execute the following from the j-ubik directory:

pytest-3 test/

Tests considering Chandra are skipped if ciao is not installed.

Guidelines for contribuation can be found in CONTRIBUTING.md

• Chandra
• eROSITA
• James Webb Space Telescope

Additional calibration files might be needed for instrument-specific pipelines.

J-UBIK allows to process observations from Chandra x-ray observatory.

• ciao >= 4.16
• marx

We recommend installation of both via conda / conda-forge ciao & marx

J-UBIK allows to process and image event files from the eROSITA x-ray observatory.

To process eROSITA observations or produce realistic synthetic data, you will need:

• eSASS, the eROSITA Science Analysis Software System. In particular, the current version of J-UBIK only supports using eSASS through the official docker container to ensure cross-compatibility.
• caldb folder, this allows to compute the eROSITA response accurately. Either the caldb from data release 1 (DR1) or from the early data release (EDR) should be present inside the data/ directory. This folder can be downloaded at caldb download.
• Download the data if you want to work with public eROSITA data, see edr and dr1.

In the demo/ repository, erosita_inference.py allows to run a generic image reconstruction with real and synthetic (mock) eROSITA data. In order to run a mock demo, you will need to download both the calibration folder as specified in the Requirements section and an actual observation, in order to build realistic exposure maps. A good example is LMC_dataset. For more information on how to run erosita_demo.py see the corresponding docstring.

J-UBIK allows to process and image event files from the James Webb Space Telescope.

In order to make use of the JWST capabilities of the package, you will need to:

• Install the jwst package.
• Install stpsf.
• Install gwcs.

For more details see jwst_demo.py in the demo/ repository. Alternatively, you can install these requirements via

pip install --user .[jwst]

NOTE:

• Importing jubik sets the floating point precision in jax to float64.
• WebbPSF has shown some compatibility issues with the numexpr package.
  The current version of the code has been tested successfully on numexpr version==2.8.4.