Mucesh, Sunil;
(2024)
Galaxy Formation and Evolution with Machine Learning: from Correlation to Causation.
Doctoral thesis (Ph.D), UCL (University College London).
Preview |
Text
Thesis.pdf - Other Download (26MB) | Preview |
Abstract
Understanding how galaxies form and evolve is at the heart of modern astronomy. With the advent of large-scale surveys, remarkable progress has been made in the last few decades as the overall picture has been established. Nevertheless, the importance of the physical processes behind the phenomena are far from known, as primarily correlations have been identified rather than the underlying causality. While simulations are inherently causal by nature, the causal effect itself is intractable given meaningful complexity. In this thesis, the causal inference framework is applied to move beyond correlations to causation, in an effort to truly understand the galaxy formation and evolution process. First—before inference (i.e., the why)—the equally important task of prediction (i.e., the what) is tackled as machine learning (ML) is utilised to predict galaxy properties. Concretely, a novel method based on the random forest (RF) algorithm is developed to generate joint probability distribution functions (PDFs). As a demonstration, joint redshift–stellar mass PDFs are estimated, which have many science applications. Compared to a traditional SED-fitting approach, the ML based method has superior performance in terms of accuracy (based on predefined metrics) and speed (by ∼ 5 orders of magnitude). Then, combining causal inference and ML, causal ML is applied to infer the causal effect of environment on galaxies, specifically on their star-formation rate (SFR). To achieve this, a comprehensive causal model of galaxy formation and evolution is constructed, and the long-outstanding problem of disentangling nature and nurture is tackled. The causal effect is found to be negative and substantial, with environment suppressing the SFR by a factor of ∼ 100. While the overall effect at z = 0 is negative, in the early Universe, environment is discovered to have a positive impact, boosting star formation by a factor of ∼ 10 at z ∼ 1 and by even greater amounts at higher redshifts.
| Type: | Thesis (Doctoral) |
|---|---|
| Qualification: | Ph.D |
| Title: | Galaxy Formation and Evolution with Machine Learning: from Correlation to Causation |
| 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/10194272 |
Archive Staff Only
 |
View Item |
