Morvan, Mario;
(2023)
Deep Learning, Shallow Dips: Transit light curves have never been so trendy.
Doctoral thesis (Ph.D), UCL (University College London).
Preview |
Text
Morvan_10163203_thesis_revised.pdf Download (14MB) | Preview |
Abstract
At the crossroad between photometry and time-domain astronomy, light curves are invaluable data objects to study distant events and sources of light even when they can not be spatially resolved. In particular, the field of exoplanet sciences has tremendously benefited from acquired stellar light curves to detect and characterise a majority of the outer worlds that we know today. Yet, their analysis is challenged by the astrophysical and instrumental noise often diluting the signals of interest. For instance, the detection of shallow dips caused by transiting exoplanets in stellar light curves typically require a precision of the order of 1 ppm to 100 ppm in units of stellar flux, and their very study directly depends upon our capacity to correct for instrumental and stellar trends. The increasing number of light curves acquired from space and ground-based telescopes—of the order of billions—opens up the possibility for global, efficient, automated processing algorithms to replace individual, parametric and hard-coded ones. Luckily, the field of deep learning is also progressing fast, revolutionising time series problems and applications. This reinforces the incentive to develop data-driven approaches hand-in-hand with existing scientific models and expertise. With the study of exoplanetary transits in focus, I developed automated approaches to learn and correct for the time-correlated noise in and across light curves. In particular, I present (i) a deep recurrent model trained via a forecasting objective to detrend individual transit light curves (e.g. from the Spitzer space telescope); (ii) the power of a Transformer-based model leveraging whole datasets of light curves (e.g. from large transit surveys) to learn the trend via a masked objective; (iii) a hybrid and flexible framework to combine neural networks with transit physics.
| Type: | Thesis (Doctoral) |
|---|---|
| Qualification: | Ph.D |
| Title: | Deep Learning, Shallow Dips: Transit light curves have never been so trendy |
| Open access status: | An open access version is available from UCL Discovery |
| Language: | English |
| Additional information: | Copyright © The Author 2023. Original content in this thesis is licensed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) Licence (https://creativecommons.org/licenses/by-nc-nd/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 |
| URI: | https://discovery-pp.ucl.ac.uk/id/eprint/10163203 |
Archive Staff Only
 |
View Item |
