Map-based emulator for CMB systematics with scattering covariances

cmbscat is a pip-installable python package that can synthesize new full-sky map samples (emulations) on the HEALPix sphere which are both visually and statistically similar to the ones found in an (eventually small) dataset of simulations. This emulator is described in detail in the accompanying paper arxiv:2503.11643.

cmbscat relies heavily on the HealpixML library for efficient scattering covariance computation on the HEALPix sphere.

Install with pip

You can install it simply with:

pip install cmbscat

We recommend using python 3.11.

Alternatively, you can install from source with:

git clone https://github.com/pcampeti/CMBSCAT.git
cd CMBSCAT/
pip install -r requirements.txt
pip install -e .

Usage

You can generate a new dataset of CMB systematics maps by doing:

from cmbscat import cmbscat_pipe

# Set emulator parameters
params = {
    'NNN'          : 10,               # Number of input reference maps
    'gauss_real'   : True,             # Generate new input data as Gaussian realizations from pixel covariance of original data
    'NGEN'         : 10,               # Batch size for gradient descent
    'n_samples'    : 10,               # Samples in the input dataset
    'nside'        : 16,               # N_side of input maps
    'NORIENT'      : 4,                # Orientations in the SC
    'nstep'        : 50,               # Steps in gradient descent
    'KERNELSZ'     : 3,                # Wavelet kernel size
    'outname'      : 'example',        # Output name
    'outpath'      : './data/',        # Output path
    'data'         : 'variable_gain_sims.npy'  # Input data path (assuming you run from inside the /CMBSCAT/notebook/ folder)
}

# Initialize pipeline...
pipeline = cmbscat_pipe(params)

#...and run! This generates NGEN new maps for each of the n_samples input maps
pipeline.run()

Tutorial Notebook

You can find an introductory notebook explaining all features of the cmbscat package here.

Specifically, we apply it to simulated maps of an example of CMB satellites instrumental systematics, as described in the accompanying paper arxiv:2503.11643.

Citations

Should this code be used in any way, we kindly ask that the following article is cited:

@article{Campeti:2025opc,
    author = "Campeti, P. and Delouis, J. -M. and Pagano, L. and Allys, E. and Lattanzi, M. and Gerbino, M.",
    title = "{From few to many maps: A fast map-level emulator for extreme augmentation of CMB systematics datasets}",
    eprint = "2503.11643",
    archivePrefix = "arXiv",
    primaryClass = "astro-ph.CO",
    month = "3",
    year = "2025"
}