Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/fbartolic/caustics
Differentiable microlensing powered by JAX
https://github.com/fbartolic/caustics
astronomy astrophysics jax
Last synced: 20 days ago
JSON representation
Differentiable microlensing powered by JAX
- Host: GitHub
- URL: https://github.com/fbartolic/caustics
- Owner: fbartolic
- License: mit
- Created: 2022-02-25T17:22:24.000Z (almost 3 years ago)
- Default Branch: main
- Last Pushed: 2022-11-26T17:14:43.000Z (about 2 years ago)
- Last Synced: 2023-04-09T14:36:51.266Z (over 1 year ago)
- Topics: astronomy, astrophysics, jax
- Language: Jupyter Notebook
- Homepage: https://fbartolic.github.io/caustics/
- Size: 3.68 MB
- Stars: 10
- Watchers: 3
- Forks: 1
- Open Issues: 19
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# caustics
[](https://github.com/fbartolic/caustics/actions/workflows/tests.yml)
`caustics` is a code for computing microlensing light curves of single, binary, and triple lens systems using the contour integration
method. It is built using the [JAX](https://github.com/google/jax) library which enables the computation
of *exact* gradients of the code outputs with respect to all input parameters through the use of [automatic differentiation](https://jax.readthedocs.io/en/latest/notebooks/autodiff_cookbook.html).
## Installation
`caustics` is still being actively developed and is not yet released on PyPI. To install the development version, clone this repository,
create a new `conda` environment, `cd` into the repository and run
```python
conda env update --file environment.yml && pip install .
```
`caustics` does not currently support Apple M1 processors except via Rosetta emulation. To install it open the terminal through Rosetta
and the command from above.
## Features
- Fast (miliseconds) and accurate computation of binary and triple lens microlensing light curves for extended limb-darkened sources.
- Automatic differentiation enables the use of gradient-based inference methods such as Hamiltonian Monte Carlo when fitting multiple lens microlensing light curves.
- A differentiable JAX version of a complex polynomial root solver [CompEA](https://github.com/trcameron/CompEA) which uses the Aberth-Ehrlich method to obtain all roots of a complex polynomial at once using an implicit deflation strategy. The gradient of the solutions with respect to the polynomial coefficients is obtained through [implicit differentiation](http://implicit-layers-tutorial.org/implicit_functions/).
- Hexadecapole approximation from [Cassan 2017](https://academic.oup.com/mnras/article/468/4/3993/3103057?login=true) is used to substantially speed up the computation of the magnification everywhere except near the caustics.
## References
- `caustics` paper coming soon!
- [Light-curve calculations for triple microlensing systems](https://academic.oup.com/mnras/article-abstract/503/4/6143/6149166?redirectedFrom=fulltext&login=false)
- [On a compensated Ehrlich-Aberth method for the accurate computation of all polynomial roots](https://hal.archives-ouvertes.fr/hal-03335604)
- [A robust and efficient method for calculating the magnification of extended sources caused by gravitational lenses](https://ui.adsabs.harvard.edu/abs/1998A%26A...333L..79D/abstract)
- [VBBINARYLENSING: a public package for microlensing light-curve computation](https://ui.adsabs.harvard.edu/abs/2018MNRAS.479.5157B/abstract)
- [Fast computation of quadrupole and hexadecapole approximations in microlensing with a single point-source evaluation](https://academic.oup.com/mnras/article/468/4/3993/3103057?login=true)