Astronomy & Astrophysics Seminar: Finding the Weirdest Things

Abstract:

Scientific discovery often hinges upon serendipity. A keen eye looking at new data (or a new problem) under a new light. However, as samples grow, this method does not scale. I will present an algorithm that we developed (Baron & Poznanski 2016), which can find outliers in large datasets. The algorithm is very general, can be applied to any large population, it is fast, and can be easily run on parallel architectures. I will show its exceptional success with galaxy spectra, identifying the weirdest galaxies in SDSS, where we find known outliers (e.g., supernovae, galaxy-galaxy lenses), but also two new classes of galaxies we did not know we should be looking for.

Seminar Organizer: Prof. Sara Beck