FGCluster is a python package aimed at running spectral clustering onto Healpix maps. The inputs are a map encoding the feature to cluster. In this case the pixel similarity is given by the geometrical affinity of each pixel in the sphere. However, if uncertainty map is provided as an input, the adjacency is modified in such a way that the pixel similarity accounts also for the statistical significance given by the pixel values in a map and the uncertainties. For further details on the methodology please see Puglisi et al. 2021 .

numpy>=1.14
matplotlib>=2.0.0
scipy>=0.19.1
astropy>=1.2
scikit-learn>=0.2.4.1
healpy>=1.10.3
mpi4py>=3.0.3

git clone https://github.com/giuspugl/fgcluster.git
cd fgcluster
python setup.py install

mpirun -np 4 python test_spectral_mpi.py

Notice that the whole package is implemented to work in parallel with mpi4py. For debugging runs please use -np 1 .

To learn how to use FGCluster we provided a tutorial notebook, available at this link

Maps of Clusters are published through Dataverse :

• PySM parameters :https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/WJUEFA
• Nclouds : https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/XAMJ4X
• GNILC dust parameters: https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/02SCHB