The use of Deep Learning in stellar classification

We are applying various ML/DL techniques for the purpose of stellar spectroscopy. Having already ran tests with Principal Component Analysis (PCA) and Sliced Inverse Regression (SIR), we now turn our focus to Convolution Neural Network (CNN), among other techniques, in order to find the most accurate derivations for stellar parameters: effective temperature, surface gravity, projected equatorial rotational velocity, microturbulence velocity and metallicity.