| Speaker: | J. Xavier Prochaska (UC Santa Cruz) |
|---|---|
| Title: | Deep Learning of Quasar Spectra |
| Date (JST): | Wed, Dec 06, 2017, 15:30 - 16:30 |
| Place: | Seminar Room A |
| Abstract: | I will describe our development of a convolutional neural network (CNN) to learn to search for and characterize absorption lines in quasar spectra. Specifically, the algorithm discovers and measures the redshift and Hydrogen column density of damped Lya systems (DLAs). These systems dominate the neutral hydrogen gas of the universe, trace the interstellar medium of distant galaxies, and offer cosmological constraints on the build up of gas and heavy elements across cosmic time. I will discuss the lessons learned employing CNN techniques on large spectral datasets and the prospects for future analysis. |
