Detection of Mg II Absorption Lines Using Convolutional Neural Networks

Nowadays, with more sky survey projects, astronomical data are generated in large quantities, and the traditional astronomical data processing methods can no longer meet the current data processing needs. Deep learning techniques are becoming more and more mature in the field of astronomy. The Mg II λ λ 2796,2803 absorption line belongs to the metallic narrow absorption line and is one of the very important absorption lines. It is widely used to study the intergalactic medium and the circumgalactic medium. In this paper, convolutional neural network is applied to Mg II absorption line detection. It is capable of multitasking and is able to classify and at the same time go on to output the redshift and equivalent width of the Mg II absorption lines. A real dataset characterized by the double peak of the Mg II absorption line was created from the published SDSS DR16 Mg II catalog. Training resulted in a model with a final accuracy of 94.5% on the test dataset, with relatively small prediction errors for equivalent width of the Mg II absorption line. The deep learning method proposed in this paper is much faster than traditional methods and maintains high accuracy with little human interference.