Using deep Residual Networks to search for galaxy-Ly α emitter lens candidates based on spectroscopic selection

Abstract

More than one hundred galaxy-scale strong gravitational lens systems have\nbeen found by searching for the emission lines coming from galaxies with\nredshifts higher than the lens galaxies. Based on this spectroscopic-selection\nmethod, we introduce the deep Residual Networks (ResNet, a kind of deep\nConvolutional Neural Networks) to search for the galaxy-Ly$\\alpha$ emitter\n(LAE) lens candidates by recognizing the Ly$\\alpha$ emission lines coming from\nhigh redshift galaxies ($2 < z < 3$) in the spectra of early-type galaxies\n(ETGs) at middle redshift ($z\\sim 0.5$). The spectra of the ETGs come from the\nData Release 12 (DR12) of the Baryon Oscillation Spectroscopic Survey (BOSS) of\nthe Sloan Digital Sky Survey \\uppercase\\expandafter{\\romannumeral3}\n(SDSS-\\uppercase\\expandafter{\\romannumeral3}). In this paper, we first build a\n28 layers ResNet model, and then artificially synthesize 150,000 training\nspectra, including 140,000 spectra without Ly$\\alpha$ lines and 10,000 ones\nwith Ly$\\alpha$ lines, to train the networks. After 20 training epochs, we\nobtain a near-perfect test accuracy at 0.9954. The corresponding loss is 0.0028\nand the completeness is 93.6\\%. We finally apply our ResNet model to our\npredictive data with 174 known lens candidates. We obtain 1232 hits including\n161 of the 174 known candidates (92.5\\% discovery rate). Apart from the hits\nfound in other works, our ResNet model also find 536 new hits. We then perform\nseveral subsequent selections on these 536 hits and present 5 most believable\nlens candidates.\n