This project provides a simulator for gravitational lensing based on Chris Clarkson's Roulettes framework. The initial prototype was an undergraduate final year project by Ingebrigtsen, Remøy, Westbø, Nedreberg, and Austnes (2022). The software includes both a GUI simulator for interactive experimentation, and a command line interface for batch generation of datasets.
The intention is to be able to use the synthetic datasets to train machine learning models which can in turn be used to map the dark matter of the Universe. This is work in progress.
Documentation is being written at https://cosmoai-aes.github.io/, but it still incomplete and fragmented.
The libraries are built using Github Workflows and packed as complete Python modules with lirbraries or scripts. Currently, we build for Python 3,9, 3.10, 3.11, and 3.12 on Linux, Python 3.10, 3.11, and 3.12 on Windows, Python 3.11 and 3.12 on MacOS for arm64, and Python 3.10 for MacOS for Intel. We only test it on linux, and it may or may not work on other platforms.
CosmoSimPy.zip from
the latest release.CosmoSimPy/CosmoGUI.py in python. This is the GUI tool.CosmoSimPy/datagen.py is the CLI tool and should be run
on the command line; see below.The GUI is also available as desktop applications for MacOS 12 on Intel and MacOS 14 on arm64. The desktop application integrates its own Python interpreter, so that you do not have to install Python separately. However, you will have to manually assure the Mac that you can trust the software.
The binaries are not signed, and you will have to confirm that you trust the binary before it will run.
The build procedure is primarily developed on Debian Bullseye, but it now also works reliable on github runners running Windows, Ubuntu, and MacOS. We have limited capacity to develop generic and robust build procedures, but we shall be happy to incorporate contributions.
The following instructions are Linux specific.
To install using conan, you can use the build script included, We require version 1.64. The setup is not compatible with conan 2.x. On some architectures version 1.59 works. You may also have to create a default profile. (See Conan Tutorial for further information.)
pip3 install conan==1.64
conan profile new default --detect
sh build.shDuring the process, it will tell you about any missing libraries that have to be installed on the system level.
If you want to save time on trial and error, you may install the following first (Debian).
sudo apt-get install libgtk2.0-dev libva-dev libx11-xcb-dev libfontenc-dev libxaw7-dev libxkbfile-dev libxmuu-dev libxpm-dev libxres-dev libxtst-dev libxvmc-dev libxcb-render-util0-dev libxcb-xkb-dev libxcb-icccm4-dev libxcb-image0-dev libxcb-keysyms1-dev libxcb-randr0-dev libxcb-shape0-dev libxcb-sync-dev libxcb-xfixes0-dev libxcb-xinerama0-dev libxcb-dri3-dev libxcb-util-dev libxcb-util0-dev libvdpau-devThe setup with conan currently works linux with superuser priviliges as of 24-04-08. Superuser privileges are required to install some dependencies using the standard package managers.
As an alternative to conan, vcpkg may be used. To use vcpkg, pass
-DCMAKE_TOOLCHAIN_FILE="[path to vcpkg]/scripts/buildsystems/vcpkg.cmake"
-DCOSMOSIM_USE_CONAN=OFFto CMake, e.g.
cmake . -B build -DCMAKE_BUILD_TYPE=Release -DCMAKE_TOOLCHAIN_FILE=../vcpkg/scripts/buildsystems/vcpkg.cmake -DCOSMOSIM_USE_CONAN=OFF As of 24-04-08 this does not work on linux. It seems to require a non-existent package.
Note: On Windows you need to copy all .dll files from
/libintoCosmoSimPy/CosmoSimin order to get the Python code to work.
Once built, an illustrative test set can be generated by the following command issued from the root directory. Note that you have to install python dependencies from the requirements file. You may want to install python libraries in a virtual environment. Here this is given to do that in global user space.
pip3 -r requirements.txt install
mkdir images
python3 CosmoSimPy/datagen.py -CR -Z 400 --csvfile Datasets/debug.csv --directory imagesThis generates a range of images in the newly created images directory. This should be on the .gitignore.
The flags may be changed; -C centres det distorted image in the centre
of the image (being debugged); -Z sets the image size; -R prints an
axes cross.
If neither conan nor vcpkg can be used, OpenCV and SymEngine must be installed
on the system. We have set up cmake not to use conan when the
hostname starts with idun, in which case the idunbuild.sh
script can be used for installation. This has been designed
for the NTNU Idun cluster, but can be tweaked for other systems.
There are two ways to interact with CosmoSim; through a GUI tool and the CLI.
python3 CosmoSimPy/CosmoGUI.pyThe GUI tool is hopefully quite self-explanatory.
The images shown are the actual source on the left and the distorted (lensed)
image on the right.
Docker images have been created to build and run the GUI. It should be possible to build and run them as follows, assuming a Unix like system.
cd docker-sim && docker build -t dockergui .
docker run -ti --rm -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix -u $(id -u):$(id -g) cosmoguiTo generate images from specified parameters, you can use
python3 CosmoSimPy/datagen.py -S sourcemodel -L lensmodel -x x -y y -s sigma -X chi -E einsteinR -n n -I imageSize -N name -R -CHere are the options specified:
lensmodel is p for point mass (exact), r for Roulette (point mass),
or s for SIS (Roulette).sourcemodel is s for sphere, e for ellipse, or t for
triangle.-C centres the image on the centre of mass (centre of light)-R draw the axes crossx and y are the coordinates of the actual sources is the standard deviation of the sourcechi is the distance to the lens in percent of the distance to the sourceeinsteinR is the Einstein radius of the lensn is the number of terms to use in roulette sum.imageSize size of output image in pixels. The image will be
imageSize$\times$imageSize pixels.name is the name of the simulation, and used to generate filenames.--help for a complete list of options.Alternatively, you can parse in a csv file to bulk generate images. Parameters which are constant for all images may be given on the command line instead of the CSV file.
python3 CosmoSimPy/datagen.py --csvfile Datasets/debug.csv --mask -R -CThe following script creates a CSV file of random parameter sets to use with the --csvfile option above.
python3 CosmoSimPy/datasetgen.pyIt should be tweaked to get the desired distribution.
The datasets generated from datasetgen.py give the parameters for the
lens and the source, as well as the image file.
This allows us to train a machine learning model to identify the lens
parameters, assuming a relatively simple lens model.
It is still a long way to go to map cluster lenses.
An alternative approach is to try to estimate the effect (lens potential)
in a neighbourhood around a point in the image. For instance, we may want
to estimate the roulette amplitudes in the centre of the image.
The datagen.py script can generate a CSV file containing these data along
with the image, as follows:
mkdir images
python3 CosmoSimPy/datagen.py -C -Z 400 --csvfile Datasets/debug.csv \
--directory images --outfile images.csv --nterms 5The images should be centred (-C); the amplitudes may not be
meaningful otherwise. The --directory flag puts images in
the given directory which must exist. The image size is given by
-Z and is square. The input and output files go without saying.
The number of terms (--nterms) is the maximum $m$ for which the
amplitudes are generated; 5 should give about 24 scalar values.
The amplitudes are labeled alpha[$s$,$m$] and beta[$s$,$m$]
in the outout CSV file. One should focus on predicting the amplitudes
for low values of $m$ first. The file also reproduces the source
parameters, and the centre of mass $(x,y)$ in the original co-ordinate
system using image coordinates with the origin in the upper left corner.
The most interesting lens model for this exercise is PsiFunctionSIS (fs), which gives the most accurate computations. The roulette amplitudes have not been implemented for any of the point mass lenses yet, and it also does not work for «SIS (rotated)» which is a legacy implementation of the roulette model with SIS and functionally equivalent to «Roulette SIS« (rs).
Warning This has yet to be tested properly.
The simulator makes numerical calculations and there will always be approximation errors.
This occurs when conan does not correctly set up the default profile. Sometimes the compiler and version fields are missing.
Run:
conan profile show defaultand check (update) these if they are wrong. Check gcc version using gcc --version.
conan profile update settings.compiler=gcc default
conan profile update settings.compiler.libcxx=libstdc++11 default
conan profile update settings.compiler.version=<gcc version e.g 11.4> defaultThese are what I used on Ubuntu, but may vary from system to system.