When: Tuesday, November 12th, 2019 @ 10:00am Where: PAB, 6th Floor, eScience Studio, Seminar Rm.
Time-domain Astrophysics in the Era of Big Data
The Large Synoptic Survey Telescope (LSST) will begin science operations in 2023 and is expected to increase the discovery rate of extragalactic transients by two orders of magnitude. With this transition comes the important question: how do we classify these events and separate the interesting “needles” from the “haystack” of objects? While it’s necessary to pick out these needles, it is equally important to understand the haystack — or the millions of events which lack any multi wavelength or spectroscopic followup.
In this talk, I will discuss ongoing efforts to classify and characterize the future haystack. In particular, I will introduce two methods to classify supernovae based on their optical light curves, which have been trained and tested on data from the Pan-STARRS Medium Deep Survey. The first method combines Bayesian model fitting with a variety of supervised classification methods, while the second method uses a semi-supervised method (a recurrent neural network-based autoencoder). I will then discuss how well we can extract physical insights from LSST-like light curves without additional followup. I will focus on the rare class of Type I superluminous supernovae as a case study, and I show that we can determine key physical properties of their engine for a significant fraction of events, indicating that population studies of rare transients will be possible with LSST.
About Ashley Villar
Ashley Villar is a Ford Foundation Dissertation Fellow at Harvard University. She studies the eruptions, mergers and explosions of stars using statistical and data-driven methods. Her 10am talk, on Tuesday, November 12th, will be on “Time-domain Astrophysics in the Era of Big Data.” Ashley completed her undergraduate degree in Physics at MIT in 2014. She is super into the statistics of ghost hunting.