The authors construct a training dataset using mock observations of galaxies from the EAGLE hydrodynamical cosmological simulation, for which the true 3D shapes are known. They measure common kinematic and photometric parameters from these mock observations and use principal component analysis to select the most informative features.
The authors then build a mixture density network (MDN) model that takes these selected features as input and outputs the probability distributions of the two axis ratios (p and q) that describe the 3D shape of each galaxy. This approach allows the MDN to capture the non-linear relationships between the observed properties and the underlying 3D shape, without making assumptions about the expected relationships.
The authors demonstrate that the MDN model can potentially improve upon previous methods in retrieving the 3D galaxy shape, especially for prolate and triaxial systems. They make recommendations for recovering galaxy intrinsic shapes relevant for current and future integral field spectroscopic galaxy surveys.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Suk Yee Yong... at arxiv.org 04-09-2024
https://arxiv.org/pdf/2404.04491.pdfDeeper Inquiries