Speaker
Description
Component separation is critical in CMB data analysis to clean foregrounds from CMB. The deep convolutional neural networks (CNN) have been increasingly useful in image segmentation problems of various fields to reconstruct the signal. I have built a hybrid CNN architecture in wavelet space (specifically in Needlet space ) to separate foregrounds from CMB that works in multi-resolution spherical data. This hybrid method introduces fewer residuals in recovered CMB than the conventional component separation techniques. I applied the trained network to Planck data. In this talk, I will discuss the robustness of the method in the recovery of CMB map irrespective of foreground complexity. The results suggest that this hybrid architecture can provide a promising alternative approach to the component separation of CMB observations.
| Would you be interested in presenting a poster if the conference is oversubcribed? | No |
|---|
