Gerardi, Francesca;
(2024)
Simulation-based inference and data compression applied to cosmological problems.
Doctoral thesis (Ph.D), UCL (University College London).
Preview |
Text
Thesis_Gerardi_final.pdf - Accepted Version Download (7MB) | Preview |
Abstract
At this time, for cosmology, it is crucial to develop robust inference frameworks that will improve our understanding of the standard cosmological model, shedding light to open problems it currently suffers of. However, the increasing complexity of current analyses urged the need for optimal data compression algorithms and alternative simulation-based Bayesian approaches, which development was boosted by Machine Learning advances. The work presented in this thesis focuses on testing and applying these techniques to problems in the field of gravitational waves and the Lyman-α forest, and the content can be effectively split up into two main building blocks. The first part aims at addressing whether density estimation simulation-based inference yields unbiased estimates of cosmological parameters in the presence of selection effects. As a test case we use mock binary neutron stars mergers catalogues for the Hubble constant estimation, given a toy hierarchical model. Not only did this method yield statistically unbiased estimates of H0 , but its precision almost matched the one of standard Bayesian analysis. The second part of my work explores if and how information can be optimally and efficiently extracted from Lyman-α correlation functions. First, we aim to understand whether the baryon acoustic oscillations peak alone, as considered in standard analyses, constitutes a sufficient summary to capture all the relevant cosmological information. Performing a direct fit to the full shape of simple mock correlations, we demonstrated that there is extra information and we traced it back to the Alcock-Paczyński effect and redshift space distortions. Finally, in another work, we apply score compression to realistic mocks, finding good agreement with the traditional approach at the posterior level. Moreover, we find that the covariance matrix estimated from data via subsampling is a good approximation to the true covariance.
| Type: | Thesis (Doctoral) |
|---|---|
| Qualification: | Ph.D |
| Title: | Simulation-based inference and data compression applied to cosmological problems |
| Open access status: | An open access version is available from UCL Discovery |
| Language: | English |
| Additional information: | Copyright © The Author 2024. Original content in this thesis is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) Licence (https://creativecommons.org/licenses/by-nc/4.0/). Any third-party copyright material present remains the property of its respective owner(s) and is licensed under its existing terms. Access may initially be restricted at the author’s request. |
| UCL classification: | UCL UCL > Provost and Vice Provost Offices > UCL BEAMS UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Maths and Physical Sciences UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Maths and Physical Sciences > Dept of Physics and Astronomy |
| URI: | https://discovery.ucl.ac.uk/id/eprint/10186141 |
Archive Staff Only
 |
View Item |
