Neural network optical characterization of accreting black holes
Astronomers have long used optical spectra (detailed wavelength-specific measurements of light) to categorize galaxies. A galaxy’s spectrum encodes information about its constituent stars, gas, and dust and its central super-massive black hole (SMBH). SMBHs that are actively accreting, known as active galactic nuclei (AGN) are of particular interest to astronomers interested in SMBH physics and their role in galaxy evolution. However, because optical spectra are influenced by so many different physical processes, their use as an indicator of AGN presence is not always reliable. Fortunately, gas that is heated by the AGN also produces x-rays which can be detected by specialized space telescopes. X-ray detections are more reliable indicators of AGN, but are also more costly to obtain. In this project, we will make use of data from the all-sky eROSITA x-ray space observatory available at the host institution, and use advanced artificial neural network (ANN) machine learning techniques to predict the presence of x-ray AGN based on the optical spectra of galaxies. Once the neural network has been trained on the optical spectra of eROSITA x-ray sources, it will be used to make predictions of AGN status in galaxies based on their optical spectra alone.
Voir la description complète du projetSara Ellison
Max Planck Institute for Extraterrestrial Physics
Physics
Artificial Intelligence
University of Victoria
Globalink Research Award
