Check out the tutorials to:
- generate synthetic images of optical galaxy cluster observations.
- reduce and process synthetic images with metacalibration to obtain the input shear field.
- end-to-end method validation environment for numerical methods in weak lensing measurements of galaxy clusters
The package is aimed to be representative of the data analysis setup of the Dark Energy Science Consorcium (DESC) of the Legace Survey of Space and Time (LSST).
Check out the Intro presentation
- artificially generated data
- trained to be representative of the real scenario
- used to test and evaluate algorithms, models, and software pipelines.
In the domain of galaxy cluster observations our aim is to measure the weak gravitational lensing signal induced by the gravitational potential wells of galaxy clusters, which enables us to estimate their masses. Validating and calibrating the weak lensing measurement is revealed to be an complex task, and the aim of this package is to provide the synthetic data which is then used for testing and statistical estimators and software pipelines.
Some illustrations of an earlier publication in our team is provided at the following link
Please see the DESIGN doc file for the detailed vision
There are three main components of this package:
Part 1: generative galaxy cluster line of sight model
This step is based on Varga et al 2021, and includes
- training the generative model (heads up: high computation intensity)
- drawing random realizations of a galaxy cluster as a member galaxy catalog
Part 2: Image rendering
- render synthetic galaxy cluster catalogs into synthetic images
Part 3: image processing and shear estimation
- find sources in the synthetic images
- prepare the image data and source catalog for further processing
- use the metacalibration algorithm to estimate the shear and response terms of the source catalog
For a first time user, please start with the tutorial notebook series in the tutorial folder.
To get the full data set, see instructions on DATA ACCESS
Right now the best way to run this package is via a pre set conda (Anaconda) environment.
At the moment this is available in the LMU USM server.
On the USM you should use the following command to activate the environment
conda activate /home/moon/vargatn/anaconda3/envs/galsim
and use the checkup script
python env_checkup.py
If this runs without errors, then you are set to go!
At the moment this package is installed from github, we are actively working to bring this to conda and pip
git clone https://github.com/vargatn/synthetic.git
then use the supplied install script
Be sure to have anaconda setup on your machine (not miniconda). We've tested the install across multiple machines with
conda 23.1.0
There's an automated install script, which you have torun in interactive mode this is important for the environment management
bash -i mamba_install.sh
The script aims to set up a new conda environment for you named synth,
with all dependencies and the synthetic package installed
you can activate it as
conda activate synth
and deactivate it as
conda deactivate
If something goes wrong, you can delete it as
conda env remove -n synth
Dependencies Key dependencies and versions are listed here
To make the end-to-end data generation, rendering and metacalibration steps work there is a fair bit of external packages which need to work together
It is our aim to provide a working ipython kernel on the DESC machines, and locally at the USM, where the package is validated to run.
TBA sphinx autodoc from docstrings
In case of suggestions, please contact me at
T.Varga @ physik.lmu.de