Deep Needlet: A CNN based full sky component separation method in Needlet space
Abstract
One of the most important steps of Cosmic Microwave Background (CMB) data analysis is component separation to recover CMB map by combining the observations contaminated by foregrounds. Needlet Internal Linear Combination (NILC) is one of the successful methods that applies the minimum variance estimation technique to a set of needlet-filtered frequency maps to recover CMB. This work develops a deep convolutional neural network (CNN) model to recover CMB map from needlet-filtered frequency maps over the full sky. The model allows to perform component separation with a multi-resolution representation of spherical data localized on both pixel space and harmonic space with rotational invariant features of CMB. The network model is trained on realistic simulations at Planck frequencies. We demonstrate the network performance for simulations that exhibit different foreground complexities. The model achieves precise recovery of the CMB temperature map and the TT power spectrum exhibits excellent agreement with true value up to $\ell\sim$1100. The residual leakage to the recovered CMB is reduced significantly compared to the CMB map recovered using NILC pipeline. Once validated on the simulations, the network is applied to Planck PR3 data to recover CMB. The recovered CMB map shows excellent agreement with CMB maps of Planck legacy products using NILC and SMICA pipelines. This work demonstrates a powerful component separation method to clean spherical signal data from multi-resolution wavelet-filtered maps.
- Publication:
-
arXiv e-prints
- Pub Date:
- January 2025
- DOI:
- arXiv:
- arXiv:2501.07469
- Bibcode:
- 2025arXiv250107469A
- Keywords:
-
- Astrophysics - Cosmology and Nongalactic Astrophysics;
- Astrophysics - Astrophysics of Galaxies
- E-Print:
- 20 pages, 11 figures, 2 tables