Deep learning of quasar spectra to discover and characterize damped Lyα systems

We have designed, developed, and applied a convolutional neural network (CNN) architecture using multi-task learning to search for and characterize strong H I Lyα absorption in quasar spectra. Without any explicit modelling of the quasar continuum or application of the predicted line profile for Lyα from quantum mechanics, our algorithm predicts the presence of strong H I absorption and estimates the corresponding redshift z_abs and H I column density N_{H I}, with emphasis on damped Lyα systems (DLAs, absorbers with N_{H I}≥ 2 × 10^{20} cm^{-2}). We tuned the CNN model using a custom training set of DLAs injected into DLA-free quasar spectra from the Sloan Digital Sky Survey (SDSS), data release 5 (DR5). Testing on a held-back validation set demonstrates a high incidence of DLAs recovered by the algorithm (97.4 per cent as DLAs and 99 per cent as an H I absorber with N_{H I}> 10^{19.5} cm^{-2}) and excellent estimates for z_abs and N_{H I}. Similar results are obtained against a human-generated survey of the SDSS DR5 data set. The algorithm yields a low incidence of false positives and negatives but is challenged by overlapping DLAs and/or very high N_{H I} systems. We have applied this CNN model to the quasar spectra of SDSS DR7 and the Baryon Oscillation Spectroscopic Survey (data release 12) and provide catalogues of 4913 and 50 969 DLAs, respectively (including 1659 and 9230 high-confidence DLAs that were previously unpublished). This work validates the application of deep learning techniques to astronomical spectra for both classification and quantitative measurements.