# How complex are galaxies? A non-parametric estimation of the intrinsic dimensionality of wide-band photometric data

**Authors:** Corentin Cadiou, Clotilde Laigle, Oscar Agertz  
**arXiv:** [2404.02962](https://arxiv.org/abs/2404.02962)  
**Date:** 2024-04-03  
**Summary page:** https://www.summarizepaper.com/en/paper/2404.02962v1/

## Key points

- Researchers Corentin Cadiou, Clotilde Laigle, and Oscar Agertz investigate galaxy complexity by determining the number of independent parameters needed to describe galaxies.
- Non-parametric method used to estimate intrinsic dimensionality of large datasets from COSMOS2020 and Horizon-AGN simulation catalogues.
- Majority of variance in wide-band photometry for star-forming and passive galaxies captured with few independent parameters.
- Star-forming galaxies require approximately 4.3±0.5 independent parameters, while passive galaxies need around 2.9±0.2 parameters.
- Noise-driven parameter identified; stellar mass and redshift key factors influencing magnitudes in observed and simulated catalogues.
- Wide-band photometry may introduce no more than one additional independent parameter for star-forming galaxies.
- Dimensionality reduction techniques effective for analyzing wide-band photometric data using at least a four-dimensional latent space.
- Research highlights importance of understanding underlying galaxy parameters for comprehensive analysis in astrophysics research.

## Summary

In this study, researchers Corentin Cadiou, Clotilde Laigle, and Oscar Agertz explore the complexity of galaxies by investigating the number of independent parameters needed to describe them. They use a non-parametric method to estimate the intrinsic dimensionality of large datasets containing wide-band photometric data from the COSMOS2020 catalogue and a mock catalogue from the Horizon-AGN simulation. The team's results indicate that despite limitations in signal-to-noise ratio across optical and NIR bands in their galaxy catalogues, the majority of variance in wide-band photometry for both star-forming and passive galaxies can be captured with only a few independent parameters. Specifically, they find that star-forming galaxies require approximately 4.3±0.5 independent parameters while passive galaxies need around 2.9±0.2 parameters to accurately describe their characteristics. Moreover, the researchers identify one parameter as noise-driven and highlight stellar mass and redshift as key factors influencing magnitudes in both observed and simulated catalogues. This suggests that wide-band photometry may not introduce more than one additional independent parameter for star-forming galaxies. While acknowledging potential limitations such as sample incompleteness due to SNR cuts, the study proposes that dimensionality reduction techniques can effectively analyze wide-band photometric data when utilizing at least a four-dimensional latent space. Overall, this research sheds light on the intricate nature of galaxies and underscores the importance of understanding their underlying parameters for comprehensive analysis and interpretation of observational data in astrophysics research.