cosmogp is a Python package designed for field-level inference of 2D weak lensing convergence fields using Gaussian processes. It includes functionalities for generating Gaussian random fields (GRF) and lognormal fields, converting between power spectra and correlation functions, and setting up numpyro models for MCMC simulations to derive cosmological parameter posteriors. The package builds upon jax-cosmo for power spectrum generation and tinygp for Gaussian processes.
-
Field Generation:
- Gaussian Random Fields (GRF)
- Lognormal Fields
-
2-pt Statistics Conversion Routines:
- FFT2 for 2D Fourier transforms
- FFTlog for Hankel transforms
- Bessel function integration for Hankel transforms
-
Cosmological Parameter Inference:
-
numpyromodel setup for MCMC simulation - Posterior distributions for cosmological parameters such as
$S_8$ ,$\Omega_m$ ,$\sigma_8$
-
git clone --recursive https://github.com/massarin/cosmogp.git
pip install -r cosmogp/requirements.txtCheck out the tutorial notebook.
@mastersthesis{Massari_Weak_lensing_map_2024,
author = {Massari, Nicolò},
doi = {10.5281/zenodo.16085961},
month = apr,
title = {{Weak lensing map inference: a physics-informed Gaussian process approach}},
year = {2024}
}