Analysis of Stellar Spectra from LAMOST DR5 with Generative Spectrum Networks

We derive the fundamental stellar atmospheric parameters (Teff, log g, [Fe/H] and [alpha/Fe]) of low-resolution spectroscopy from LAMOST DR5 with Generative Spectrum Networks (GSN), which follows the same scheme as a normal ANN with stellar parameters as the inputs and spectrum as outputs. After training on PHOENIX theoretical spectra, the GSN model performed effectively on producing synthetic spectra. Combining with a Bayes framework, application in analysis of LAMOST observed spectra becomes efficient on the Spark platform. Also, we examined and validated the results by comparing with reference parameters of high-resolution surveys and asteroseismic results. Our method is credible with a precision of 130K for Teff, 0.15 dex for log g, 0.13 dex for [Fe/H] and 0.10 dex for [alpha/Fe].