The Survey of Surveys: machine learning for stellar parametrization

A. Turchi,E. Pancino,F. Rossi,A. Avdeeva,P. Marrese,S. Marinoni,N. Sanna,M. Tsantaki,G. Fanari

DOI: https://doi.org/10.1117/12.3018967

2024-12-06

Abstract:We present a machine learning method to assign stellar parameters (temperature, surface gravity, metallicity) to the photometric data of large photometric surveys such as SDSS and SKYMAPPER. The method makes use of our previous effort in homogenizing and recalibrating spectroscopic data from surveys like APOGEE, GALAH, or LAMOST into a single catalog, which is used to inform a neural network. We obtain spectroscopic-quality parameters for millions of stars that have only been observed photometrically. The typical uncertainties are of the order of 100K in temperature, 0.1 dex in surface gravity, and 0.1 dex in metallicity and the method performs well down to low metallicity, were obtaining reliable results is known to be difficult.

Instrumentation and Methods for Astrophysics,Astrophysics of Galaxies,Solar and Stellar Astrophysics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: How to use machine - learning methods to derive the key parameters (effective temperature \(T_{\text{eff}}\), surface gravity \(\log g\) and metallicity \([Fe/H]\)) of stars from the data of large - scale photometric surveys (such as SDSS and SkyMapper), so as to provide parameter estimates of spectral - like quality for millions of stars observed only by photometric measurements.

### Specific problems:
1. **Large amount of data and difficulty in high - quality spectral measurement**: Although large - scale spectral surveys (such as APOGEE, GALAH, LAMOST, etc.) provide high - quality spectral data for a large number of stars, obtaining these data requires a large amount of telescope time, so it is difficult to cover all star samples. And although photometric data is huge in quantity, its precision is low.

2. **Difficulties in parameter estimation for stars with low metallicity**: In the case of low metallicity, it is especially difficult to obtain reliable star parameters.

### Solutions:
The paper proposes a machine - learning - based method. Using previous work on homogenizing and recalibrating spectral data, a neural network model is constructed. This model can derive high - precision star parameters from photometric data. Specific steps include:
- **Data integration**: Cross - match photometric data such as Gaia and SkyMapper with spectral data such as APOGEE and GALAH to form a high - quality input data set containing tens of millions of stars.
- **Neural network training**: Use the spectral parameters in the SoS (Survey of Surveys) catalog as "true" values to train the neural network to predict \(T_{\text{eff}}\), \(\log g\) and \([Fe/H]\) of stars.
- **Performance evaluation**: Verify the accuracy of the model through the test set. The results show that the errors of the model on the test set are: the standard deviation of \(T_{\text{eff}}\) is 85 K, the standard deviation of \(\log g\) is 0.14 dex, and the standard deviation of \([Fe/H]\) is 0.12 dex.

### Significance:
This method enables researchers to use the existing large - scale photometric data to provide high - quality parameter estimates for hundreds of millions or even billions of stars, greatly expanding the available high - quality star parameter samples, and helping astronomers to study the structure and evolution of the Milky Way more in - depth.