Tohill, Clár-Bríd
(2024)
Exploring the Morphologies of High-Redshift Galaxies with Deep Learning.
PhD thesis, University of Nottingham.
Abstract
This thesis explores different machine learning techniques for the study of galaxy morphology, morphological classification, and the morphological evolution of galaxies. We utilise data from all of the CANDELS fields imaged with HST, as well as data from the CEERS program imaged with JWST.
In Chapter 2 we train convolutional neural networks to predict the non-parametric, concentration and asymmetry measurements from individual images of ∼150, 000 galaxies at 0 < z < 7 7 in the CANDELS fields. We apply a Bayesian hyperparameter optimisation to select suitable network architectures for this problem. Our resulting networks accurately reproduce measurements compared with standard algorithms. Furthermore, using simulated images, we show that our networks are more stable than the standard algorithms at low signal-to-noise. While both approaches suffer from similar systematic biases with redshift, these remain small out to z ∼ 7. Once trained, measurements with
our networks are > 103 times faster than previous methods. Our approach is thus able to reproduce standard measures of non-parametric morphologies and shows the potential of employing neural networks to provide superior results in substantially less time. This will be vital for making best use of the large and complex datasets provided by upcoming galaxy surveys, such as Euclid and Rubin-LSST.
In Chapter 3, we employ variational auto-encoders to perform feature extraction on galaxies at z < 2 using JWST/NIRCam data. Our sample comprises 6869 galaxies at z < 2, including 255 galaxies with z < 5 (when the Universe was ∼ 9% of its current age), which have been detected in both the CANDELS/HST fields and CEERS/JWST, ensuring reliable measurements of redshift, mass, and star formation rates. To address potential biases, we eliminate background sources within our galaxy images, allowing our model to focus on the target galaxy. In order to avoid complicating the learned feature space, we eliminate galaxy orientation prior to encoding the galaxy features. We also normalise the apparent size of our sources prior to feature extraction, thereby constructing a physically meaningful feature space. By clustering the resulting feature space, we identify 11 distinct morphological classes that exhibit clear separation in various structural parameters, such as CAS-M20, Sérsic indices, specific star formation rates, and axis ratios. We observe a decline in the presence of spheroidal-type galaxies with increasing redshift, indicating a dominance of disk-like galaxies in the early universe. We demonstrate that conventional visual classification systems are inadequate for high-redshift morphology classification and advocate the need for a more detailed and refined classification scheme. Leveraging machine-extracted features, we propose a solution to this challenge and illustrate how our extracted clusters align with measured parameters, offering greater physical relevance compared to traditional methods.
In Chapter 4, we expand upon the work carried out in Chapter 3 by incorporating the redshift information of our galaxy sample into our network. By conditioning our network on the redshift of our galaxy sample, we allow the network to encode a more efficient representation of our dataset, whilst simultaneously creating a tool to explore morphological variations with redshift. We introduce a novel technique of ß-decay to our network, which forces the network to make good use of the redshift condition, whilst offering a method to determine the dominant features within a dataset. This work is on-going and we conclude with the next steps that this project will entail.
We finish with a brief overview of each research project, along with how the work in this thesis can be beneficial to future ‘Big Data’ surveys. We detail some improvements that can be made to this and other works, to further the study of galaxy morphology.
| Item Type: |
Thesis (University of Nottingham only)
(PhD)
|
| Supervisors: |
Bamford, Steven
Conselice, Christopher |
| Keywords: |
machine learning, galaxies, redshift, morphological classification, morphological evolution |
| Subjects: |
Q Science > QB Astronomy |
| Faculties/Schools: |
UK Campuses > Faculty of Science > School of Physics and Astronomy |
| Item ID: |
79446 |
| Depositing User: |
Tohill, Clar-Brid
|
| Date Deposited: |
13 Dec 2024 04:40 |
| Last Modified: |
13 Dec 2024 04:40 |
| URI: |
https://eprints.nottingham.ac.uk/id/eprint/79446 |
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
