Stellar Spectra Interpolation Using Machine Learning Techniques

In this study, empirical stellar spectral libraries are used to interpolate stellar spectra based on three stellar atmospheric parameters. However, the irregular distribution of stars’ data in these libraries often leads to inadequate results with traditional interpolation methods. After examining a radial basis function (RBF) network and its structure, a fully connected artificial neural network (ANN) model and a random forest (RF) model are trained to generate a stellar spectrum for a given set of atmospheric parameters. Comparisons reveal that the ANN and RF models significantly outperform RBF and TGM۲ interpolations, with both offering comparable performance and an R² score of approximately ۹۶ percent. A brief analysis shows that these machine learning models are robust against small input errors.

کلیدواژه ها:

Stellar Spectral Library ، Radial Basis Function Network ، Artificial Neural Network ، Random Forest