Applications of Machine Learning in Stellar Astrophysics
Wed 12 December, 2018 @12:00 PM, level 7
Dr George Angelou, MPA
Astronomy is now very much a big-data science. Gaia has had its second data release and is on track to measure the brightness, position and kinematics of close to 109 stars. TESS and PLATO will observe pulsations in > 105 targets. They will add to the approximately 2×105 that have had their oscillations monitored by Kepler and CoRoT. To exploit such large data sets the tools of analysis devised must demonstrate speed, accuracy and versatility. Speed and accuracy are required to process the sheer volume of data collected. Versatility is necessary because although there will be overlap in the many surveys, not all stars will have the same quantities measured. It is paramount that methods are designed to handle missing or inhomogeneous data sets — we must maximize the information extracted from the available observations. I will discuss the Stellar Parameters in an Instant Pipeline (SPI) which is a machine learning algorithm that makes use of detailed asteroseismic observations to rapidly and robustly determine stellar parameters. Stellar parameters are important for both exoplanet and galactic astrophysics.