Recovering the 3D Shape of Galaxies from Kinematic and Photometric Observations using Mixture Density Networks

Core Concepts

This work aims to recover the intrinsic 3D shape of individual galaxies using their projected stellar kinematic and flux distributions through a supervised machine learning approach with mixture density networks.

Abstract

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.

Stats

The authors measure the following key metrics from the mock observations:

Ellipticity (ϵ) and its error (Δϵ)
Sérsic index (n) and its error (Δn)
Kinematic misalignment angle (ψ) and its error (Δψ)
Observable spin parameter (λR)
Kinematic support (V/σ)
Specific angular momentum (j)
Mass-weighted velocity dispersion (σm)
Rotational velocity (Vrot)

Quotes

"Using a variety of photometric and kinematic measurements commonly derived for IFS observations of galaxies, we can probe the shape information locked into various parameters, and quantify their shape-determining "power" relative to one another."
"We are free from making such assumptions when using a non-linear, machine learning approach."

Key Insights Distilled From

Galaxy 3D Shape Recovery using Mixture Density Network

by Suk Yee Yong... at arxiv.org 04-09-2024

https://arxiv.org/pdf/2404.04491.pdf

Deeper Inquiries

How would the performance of the MDN model change if the training data included a wider range of galaxy properties, such as stellar mass, environment, or merger history

Including a wider range of galaxy properties in the training data for the MDN model could potentially improve its performance by providing a more comprehensive understanding of the factors influencing galaxy shapes. Stellar mass, for example, plays a crucial role in shaping galaxies, with more massive galaxies often exhibiting different structural properties compared to less massive ones. By incorporating stellar mass data into the training set, the MDN model could learn to better differentiate between galaxies of varying masses and their corresponding shapes.
Additionally, considering environmental factors such as galaxy clustering or proximity to other galaxies could offer insights into how the surrounding environment influences galaxy shapes. Galaxies in dense environments may experience more interactions and mergers, leading to distinct shapes compared to isolated galaxies. By including environmental data in the training set, the MDN model could potentially capture these effects and improve its shape recovery predictions.
Furthermore, incorporating information about the merger history of galaxies could enhance the MDN model's ability to distinguish between galaxies that have undergone significant interactions and those that have evolved in isolation. Galaxies with recent merger events may exhibit unique shapes and kinematic properties that can be learned by the model with the inclusion of merger history data.
In summary, expanding the range of galaxy properties in the training data, such as stellar mass, environment, and merger history, could enrich the MDN model's understanding of the diverse factors influencing galaxy shapes and lead to more accurate shape recovery predictions.

What are the potential limitations of the MDN approach, and how could it be further improved to better capture the complexities of galaxy shapes

While the MDN approach offers a powerful framework for recovering 3D galaxy shapes from 2D observations, there are potential limitations and areas for improvement.
One limitation is the assumption of an ellipsoidal shape for galaxies, which may not accurately represent the true diversity of galaxy shapes in the universe. Galaxies can exhibit complex structures and irregular morphologies that go beyond simple ellipsoids. To address this limitation, the MDN model could be enhanced by incorporating more sophisticated shape models that allow for greater flexibility in capturing the complexities of galaxy shapes.
Another potential limitation is the reliance on mock observations from simulations, which may not fully capture the complexities of real observational data. Incorporating observational data from actual galaxy surveys could improve the model's performance by providing more realistic and diverse training examples.
To further improve the MDN approach, one could explore the integration of additional data sources, such as multi-wavelength observations or spectroscopic data, to enhance the model's ability to recover 3D galaxy shapes. By incorporating a broader range of observational data, the model could gain a more comprehensive understanding of the physical processes shaping galaxies and improve its shape recovery accuracy.
Additionally, refining the architecture of the MDN model, optimizing hyperparameters, and implementing advanced regularization techniques could help mitigate overfitting and enhance the model's generalization capabilities. Continuous refinement and validation of the model on diverse datasets would be essential to ensure its robustness and reliability in recovering galaxy shapes.

Could the insights gained from this work on recovering 3D galaxy shapes be applied to other astrophysical problems involving the inference of 3D properties from 2D observations

The insights gained from the work on recovering 3D galaxy shapes using the MDN approach have broader implications for other astrophysical problems involving the inference of 3D properties from 2D observations.
One potential application is in the field of galaxy evolution, where understanding the 3D shapes of galaxies can provide valuable information about their formation and evolutionary histories. By applying similar machine learning techniques to infer 3D properties from 2D observations in galaxy evolution studies, researchers can gain deeper insights into the processes driving the diversity of galaxy shapes over cosmic time.
Furthermore, the methodology developed for recovering galaxy shapes could be extended to other astronomical objects, such as galaxy clusters, stellar systems, or even cosmological structures. By adapting the MDN approach to analyze 2D observations of these objects and infer their intrinsic 3D properties, researchers can uncover new insights into the spatial distributions, dynamics, and interactions of various astrophysical systems.
Overall, the techniques and methodologies developed for recovering 3D galaxy shapes using machine learning have the potential to revolutionize the way astronomers extract and interpret complex 3D information from 2D observational data across a wide range of astrophysical studies.