Da Eun Kang (Korea)
daeun.astro @ gmail.com
Machine learning and emission-line diagnostics in astrophysics
Young massive stars play an important role in the evolution of Interstellar medium (ISM) and self-regulation of star formation in Giant Molecular Cloud (GMC). Energy injected by massive stars via stellar wind, radiation and supernova explosion disrupts the surrounding environment. The interaction between stars and star-forming cloud, so-called stellar feedback, can destroy star-forming cloud or generate new stars when gravity is strong enough to make cloud re-collapse. Information on young stellar clusters, such as their evolution and distribution of surrounding gases, inheres in emission lines observed from star-forming regions. However, it is not easy to understand the nature of them from observation, as observed lights are results of complicated interactions between photons and gases.
To overcome the irresolvable relation between observation and nature, we are going to present a new machine learning tool that can characterize and analyze the young stellar clusters. We apply Invertible Neural Network (INN), one of deep learning architecture, to link the physical properties of young stellar clusters and emission-line luminosities. The methodology of this project consists of 1) generating the mock database using numerical simulations (i.e., WARPFIELD, CLOUDY, and POLARIS), 2) designing a proper neural network model and training it with the database, and 3) applying the network to find probability distributions of physical properties. The outcome of this project is going to be utilized to numerous spectral data from large survey projects and allow us to study the young stellar clusters with unprecedented details.
Supervisor: Simon Glover (ITA)