Speaker
Daniel de Andres
(Universidad Autonoma de Madrid)
Description
In this work, we evaluate for the first time Convolutional Neural Networks(CNNs) to infer the masses of observed galaxy clusters in the Planck Compton parameter maps. We train our network using simulated maps from the THREE HUNDRED SIMULATION project up to redshiftz of order 1 and test our model on real Planck Sunyaev-Zel’dovich (SZ) maps. Our data set consists on 191862 mock maps, which are based on 7106 different clusters from our simulations, and 1094 observed SZ maps. Furthermore, we train 4 separate CNNs for different redshifts intervals between z=0 and z=1. We show that our results are compatible with Planck estimates of the mass and also with weak lensing measurements
Primary authors
Daniel de Andres
(Universidad Autonoma de Madrid)
Marco De Petris
(ROMA1)
Florian Ruppin
(MIT)
Weiguang Cui
Prof.
Gustavo Yepes
(UAM)
