Abstract
We present a novel deep learning model, intelligent quasar continuum neural
network (iQNet), which predicts the intrinsic continuum of any quasar in the
rest-frame wavelength range 1020 Angstroms $łeq$ 1600 Angstroms.
We train this network using quasar spectra at low redshift ($z 0.2$) from
the Hubble Spectroscopic Legacy Archive, and apply it to predict quasar
continua from different astronomical surveys. We introduce a standardization
process to the data, reducing the absolute fractional flux error (AFFE) of the
predicted continua approximately by half. We use principal component analysis
and Gaussian mixture model to classify the HSLA quasar spectra into four
classes and use them to synthesize mock quasar spectra create a training data
set for iQNet. iQNet achieves a median AFFE of 1.31% on the training quasar
spectra, approximately ten times better than traditional PCA-based prediction
methods, and 4.17% on the testing quasar spectra. We apply iQNet to predict the
continua of $\sim$ 3200 quasar spectra at higher redshift ($2< z 5$) and
measure the redshift evolution of mean transmitted flux ($< F >$) in the
Ly-$\alpha$ forest region. We measure a gradual evolution of $< F >$ with
redshift, which we characterize as a power-law evolution. These estimates are
broadly consistent with literature, but provide a more accurate measurement as
we measure the quasar continuum with minimum contamination from the Ly-$\alpha$
forest. This work proves that the iQNet model can predict the quasar continuum
with high accuracy and shows the viability of such methods for quasar continuum
prediction.
Users
Please
log in to take part in the discussion (add own reviews or comments).