aboucaud / MICMAC

Pixel implementation for the non-parametric component separation

Minimally Informed CMB MAp Constructor: MICMAC [Release soon]

Pixel implementation for the non-parametric component separation. Extension to component separation method of Leloup et al. 2023 (https://journals.aps.org/prd/abstract/10.1103/PhysRevD.108.123547)

Installation

Create a clean virtual environment with the reuired dependencies:

conda env create -f micmac_env.yml

Install the micmac package by running:

cd micmac
python -m pip install .

You can then use the package by importing it as:

import micmac

Required dependencies

List of required dependencies:

• camb
• healpy
• jax
• jax-tqdm
• jax-healpy
• jaxopt
• numpyro
• lineax
• chex

Install jax as:

pip install --upgrade pip
pip install --upgrade "jax[cpu]"

Optional (for tutorials):

• cmbdb (https://github.com/dpole/cmbdb/tree/master)
• fgbuster (only uses fgbuster.observation_helpers to get the input frequency maps in the tutorials, the src code is completely independent from fgbuster)

Note DO NOT USE THIS PACKAGE ON LOGIN NODES ON HPC !!! (do not perform any computation on login-node in general, but this package in particular might use all the login node resources available)