CMU/Pitt AstroLunch: Matthew Ho (Institute for Advanced Study)

Speaker: Matthew Ho (Institute for Advanced Study), PhD from CMU in 2022 Title: Accelerating Cosmological Discovery with Trustworthy Machine Learning When: Friday, August 28 at 11am Where: McWilliams Center, Wean Hall 8330 Remote: Join via Zoom Lunch will be provided. Please use the schedule to sign up to meet with the speaker and/or join for dinner. Abstract: Modern cosmological surveys are mapping the universe at an unprecedented scale and resolution, offering a stringent testbed to probe the physics of dark matter, dark energy, and structure formation. While machine learning (ML) offers powerful tools to process these massive datasets, their black box nature risks losing scientific rigor. In this talk, I will demonstrate how ML can bridge the gap between theory and observation without sacrificing statistical reliability. I will first show how ML enhances parameter constraints from raw survey data, highlighting applications from weighing galaxy clusters to detecting rare cosmic-ray signatures. I will then present generative emulators that scale high-fidelity simulations to full survey volumes, enabling the first end-to-end ML analysis of the full Sloan Digital Sky Survey CMASS dataset. Finally, I will present interpretability techniques that check whether these models are learning genuine physical features, ultimately using these learned representations to identify gaps in existing theory and guide the search for new physics. Zoom: https://cmu.zoom.us/j/95641222956?pwd=dnJjn87S5rK5sSftFTfIRdbUGvKh4s.1&jst=2