Skip to main content
SearchLogin or Signup

Classification of Lyman-Alpha Emitters in HETDEX Using Unsupervised Machine Learning

Presentation #135.06 in the session “Intergalactic Medium”.

Published onJan 11, 2021
Classification of Lyman-Alpha Emitters in HETDEX Using Unsupervised Machine Learning

The HETDEX project will constrain the evolution of dark energy at z = 2-3 by measuring the characteristic scale-length of galaxy-galaxy separations across the cosmos (i.e. Baryon Acoustic Oscillations; BAO). Lyman-⍺ emitting (LAE) galaxies are used as tracers of these oscillations. The project uses VIRUS IFU spectrographs on the 10-m Hobby-Eberly Telescope (HET) to conduct an unbiased spectroscopic survey of two large (100 square degrees) regions in the spring and fall skies, and expects to identify 5 million emission-line objects. We present a new method to classify emission line objects as either LAE galaxies or non-LAE. We identify them with a cluster analysis of the emission-line spectra using the unsupervised machine learning algorithm t-SNE. Our results show significant separation in an abstract parameter space between the spectra of LAE galaxies and non-LAEs. We develop a quantitative method to assign a likelihood that 700 spectra were emitted by LAE galaxies. Our pipeline will ease the workload of the HETDEX BAO survey by many person-hours, and we encourage its use in other upcoming spectroscopic galaxy surveys.


Comments
0
comment

No comments here