Unsupervised Learning for Stellar Spectra with Deep Normalizing Flows - Jo Ciuca

Explore unsupervised learning techniques for analyzing stellar spectra using deep normalizing flows in this 28-minute conference talk by Jo Ciuca from the Australian National University. Dive into the application of astrostatistics and machine learning tools in galaxy formation and evolution, focusing on the analysis of vast datasets from Integral Field Unit surveys. Discover how data science tools can link observations with theoretical models and detect anomalous galaxies. Learn about the project overview, including the use of synthetic spectra, neural networks, and correlation matrices to find new spectral lines, identify blended lines, and detect unique objects and stars. Gain insights into the potential of these techniques for maximizing the understanding of galaxy formation physics and translating data-driven results into physical understanding.

Introduction
Jo Ciuca
Galaxy evolution
Big data
Stereo Spectra
Synthetic Spectra
Nonsupervised ML
Neural network
Project overview
Science
Thesis
Training
Correlation Matrix
Finding new lines
Blended lines
Outliers
Unique objects
Unique stars
Summary
Questions