35th Annual Student Research Conference
Strangers on the Moon: Detecting Anomalous Objects in the Second HETDEX Data Release with the Isolation Forest (A Machine Learning Approach)
Lauren Hoffman
Dr. Rasanjali Jayathissa, Faculty Mentor
The Hobby-Eberly Telescope Dark Energy Experiment (HETDEX) is designed to map the positions of more than one million galaxies to constrain dark energy parameters. HETDEX uniquely collects millions of spectra blindly, many of which are useful for auxiliary science projects. The project uses 69,096 spectra from the second HETDEX data release continuum catalog to search for spectra of interesting astronomical objects that are anomalous relative to the majority classes within astronomy, for example supernovae. We perform anomaly detection using an isolation forest which uses decision trees to isolate each spectrum. Once isolated, spectra that require shorter branches to isolate are considered anomalous. We visually inspect each of the spectra labeled as anomalous to determine if the spectrum is a true anomaly. We found 4,096 anomalous spectra from the total, including an anomalous Broad Absorption Line Active Galactic Nuclei (BAL AGN).
Keywords: Supernovae, Machine Learning, AGN, Isolation Forest, HETDEX, Anomalies
Topic(s):Physics
Astronomy
Presentation Type: Oral Presentation
Session: 207-1
Location: MG 1098
Time: 10:15