Day 1443

Monday.  Tuesday.

1807.08249
Cosmological parameters from weak cosmological lensing
Kilbinger

In this manuscript of the habilitation à dirge des recherchen (HDR), the author presents some of his work over the last 10 years.  The main topic of this thesis is cosmic shear, the distortion of images of distant galaxies due to WL by the LSS in the Universe.  Over the last years, cosmic shear has evolved into a reliable and robust cosmological probe, providing measurements of the expansion history of the Universe and the growth of its structure.  Review the principles of WL and show how cosmic shear is interpreted in a cosmological context.  Then, give an overview of WL measurements, and present observational results from CFHTLenS, as well as the implications for cosmology.  Conclude with an outlook on the various future surveys and missions, for which cosmic shear is one of the main science drivers, and discuss promising new weak cosmological lensing techniques for future observations.  (also see 1411.0115)


1807.08732
Cosmological constraints from noisy convergence maps through deep learning
Fluri, Kacprzak, Lucci, Refregier, Amara, Hofmann

Deep learning is a powerful analysis technique that has recently been proposed as a method to constrain cosmo parameters from WL mass maps.  Due to its ability to learn relevant features from the data, it is able to extract more information from the mass maps than the commonly used PS, and thus achieve better precision for cosmo parameter measurement.  Explore the advantage of Convolutional Neural Networks (CNN) over the power spectrum for varying levels of shape noise and different smoothing scales applied to the maps.  Compare the cosmo constraints from the two methods in the Omega_m-sigma_8 plane for sets of 400 deg^2 convergence maps.  Find that, for a shape noise level corresponding to 8.53 galaxies/arcmin^2 and the smoothing scale of sigma_s=2.34 arcmin, the network is able to generate 45% tighter constraints.  For smaller smoothing scale of sigma_s=1.17 the improvement can reach ~50%, while for larger smoothing scale of sigma_s=5.85, the improvement decreases to 19%.  The advantage generally decreases when the noise level and smoothing scales increase.  Present a new training strategy to train the neural network with noise data, as well as considerations for practical applications of the deep learning approach.