A deep learning approach to quasar continuum prediction

We present a novel intelligent quasar continuum neural network (iQNet), predicting the intrinsic of any in rest-frame wavelength range 1020 Angstroms $\leq \lambda \leq$ 1600 Angstroms. train this using high-resolution Hubble Space Telescope/Cosmic Origin Spectrograph ultraviolet spectra at low redshift ($z \sim 0.2$) from Spectroscopic Legacy Archive, and apply it to predict continua different astronomical surveys. utilize HSLA that are well-defined [1020, 1600] with an overall median signal-to-noise ratio least five. The iQNet achieves AFFE 2.24% on training spectra, 4.17% testing spectra. $\sim$3200 SDSS-DR16 higher ($2< z \leq 5$) measure evolution mean transmitted flux ($< F >$) Ly-$\alpha$ forest region. gradual $< >$ redshift, which we characterize as power-law fit effective optical depth forest. Our measurements broadly consistent other estimates $$ literature, but provide more accurate measurement directly measuring where there is minimum contamination This work proves deep learning model can high accuracy shows viability such methods for prediction.