Systematic Uncertainties in Normalizing Flows 2024

Resources: Paper (Link) | Poster (Link)

An exploratory project examining how data complexity impacts uncertainties in normalizing flow models applied to astrophysical distributions. Using toy examples, we investigated when initialization variance or data variance dominate the modeling process. This early-stage work informed our understanding of systematic uncertainties in machine learning applications to physics and was presented at the ML for Physical Sciences Workshop at NeurIPS 2024.