Skip to main content

CIERA Astronomers Use Machine Learning to Classify 1.5 Billion Astronomical Sources

The Zwicky Transient Facility (ZTF) is a new experiment attempting to identify everything that changes in the Northern night sky. ZTF measures these changes by repeatedly observing the same patch of sky every night to look for stellar explosions, known as supernovae, variable stars, and comets and asteroids. Once ZTF discovers an astronomically varying object, a critical challenge for ZTF is to determine whether the object is in our solar system (asteroids), the Milky Way galaxy (variable stars), or some other distant galaxy (supernovae).

To address this issue, a group of ZTF researchers, led by CIERA astronomer Adam Miller, built a machine learning model to classify ~1.5 billion sources measured by the Panoramic Survey Telescope and Rapid Response System (Pan-STARRS). By classifying those sources as either stars or galaxies, astronomers can immediately associate newly discovered variability with sources within or outside the Milky Way in order to better understand their nature at the moment of discovery. This is an essential step in maximizing the efficiency of ZTF and similar future surveys, like the Large Synoptic Survey Telescope (LSST; Northwestern/CIERA is a member institution of LSST).


Learn more:

Zwicky Transient Facility

Read the press release: Zwicky Transient Facility Nabs Several Supernovae a Night

Read the paper: A Morphological Classification Model to Identify Unresolved PanSTARRS1 Sources: Application in the ZTF Real-time Pipeline

The catalog is available here: