Lunch talk on Mar. 4, 2024
Using Machine Learning to Enhance the HETDEX survey
Speaker: Lindsay House (University of Texas at Austin)
Venue: Video Conference
Time: 12:30 PM, Monday, Mar. 4, 2024
Abstract:
In the age of large data astronomy, the use of machine learning (ML) is becoming more imperative. One method to ensure accurate results with ML is to use visual vetting. Dark Energy Explorers is an online campaign, hosted on Zooniverse. To date, we have reached more than 16,000 individuals, representing over 80 countries. The goal is to visually vet sources from the Hobby-Eberly Telescope Dark Energy Experiment (HETDEX). HETDEX is working to understand dark energy by mapping out millions of galaxies that are at a lookback time of about 11 billion years. As an unbiased, spectroscopic survey one of the largest issues of HETDEX to tackle is minimizing false positives and contamination of other sources. Dark Energy Explorers has been successful with training the public to differentiate galaxies from contamination to an accuracy of 98% as compared to an expert (House et al. 2023). In turn, Dark Energy Explorers has improved the science goals of HETDEX in combination with machine learning algorithms. Dark Energy Explorers and ML efforts have been successful in eliminating false positives from the source catalog (House et al. 2023) and have been able to remove almost 8,000 false detections with these efforts. Dark Energy Explorers has also used public engagement efforts to educate the public on astronomy and ML by hosting zoom nights, telescope tours, paper discussions and developing worksheets. This talk will further explain how Dark Energy Explorers and ML have been successful in improving the science goals of HETDEX and educating the public.
Report PPT: SWIFAR_Lindsay House.pdf