María Belén Costanza
Neural network developing for Wiener filter process applied to CMB
polarization maps and power spectrum computation.
In this talk I will present the neural network developed to simulate the Wiener Filter (WF) for polarization Cosmic Microwave Background (CMB) maps, used to reduce the inhomogeneous noise present in those maps and reconstruct the B-mode signal, with an additional procedure for the E-to-B leakage problem. We present the performance of this neural network, written in Tensorflow 2, comparing the results with the traditional method that uses conjugate gradient (CG). Furthermore, we estimate the power spectrum of an unknown signal and study the accuracy on different scales of interest.
Event Timeslots (1)
Tuesday
-
PDF
Neural network developing for Wiener filter process applied to CMB polarization maps and power spectrum computation
