The cosmic web: from identification to cosmological parameters
Abstract
The spatial distribution of matter at late time in the Universe depicts a complex pattern commonly referred to as ‘the cosmic web’. In this web-like structure, massive nodes are linked together by elongated bridges of matter, the filaments, themselves found at the intersections of mildly-dense walls forming the borders of vast and underdense volumes called voids. In a first part of the presentation I will introduce a method used to identify the most prominent structure of the web, the filamentary pattern. The method is based on a regularised version of Gaussian mixture models in which a spatial graph is used to model the underlying one-dimensional structure and provides a smooth estimate of a graph passing in the middle of the galaxy distribution. In a second part of the talk, I will show how to use the several environments (filaments, but also voids, walls, and nodes) to improve the constraints on cosmological parameters over the traditionally-used two-point statistics. In particular, by breaking some degeneracies between parameters of the model, we can improve by up to an order of magnitude the constraints on some parameters. Finally, if time permits, I will also take a few minutes to discuss an other part of my recent research activity aiming at using developments from theoretical physics to better understand the learning procedure of some simple neural networks.
