Inpainting CMB maps using partial convolutional neural networks

Abstract

We present a novel application of partial convolutional neural networks (PCNN) that can inpaint masked images of the cosmic microwave background. The network can reconstruct both the maps and the power spectra to a few percent for circular and irregularly shaped masks covering up to 10% of the image area. By performing a Kolmogorov-Smirnov test we show that the reconstructed maps and power spectra are indistinguishable from the input maps and power spectra at the 99.9% level. Moreover, we show that PCNNs can inpaint maps with regular and irregular masks to the same accuracy. This should be particularly beneficial to inpaint irregular masks for the CMB that come from astrophysical sources such as Galactic foregrounds. The proof of concept application shown in this paper shows that PCNNs can be an important tool in data analysis pipelines in cosmology.

Machine Learning for Observational Cosmology

“…Many realizations of wide-area, high-resolution maps generated by ML methods can be used to study potential systematic errors and to evaluate covariance matrices, which are crucial for precise cosmological analysis. The use of conditional generative models are also proposed for other purposes such as removing the foreground [172], separation of each component [173], reconstruction of lensing map [174] and in-painting of masked regions [175][176][177][178]. To utilize all-sky data, one can extend a traditional 2D CNN to be applied to images on a sphere.…”

Machine learning for observational cosmology

“…Many realizations of wide-area, high-resolution maps generated by ML methods can be used to study potential systematic errors and to evaluate covariance matrices, which are crucial for precise cosmological analysis. The use of conditional generative models are also proposed for other purposes such as removing the foreground [172], separation of each component [173], reconstruction of lensing map [174] and in-painting of masked regions [175][176][177][178]. To utilize all-sky data, one can extend a traditional 2D CNN to be applied to images on a sphere.…”

Cosmological particle production and pairwise hotspots on the CMB

“…-cutting the sky into 'flattened' patches -is not technically necessary but is often adopted in the literature [42][43][44][45][46]. Historically it was used since CMB analyzers using wavelets or machine learning could not cope with spherical coordinates [37][38][39][47][48][49]. This technicality has been overcome in the neutral network analysis [50,51], but we will stick to the flat patches for simplicity in this analysis.…”