Francisco Villaescusa-Navarro (Flatiron Institute) – AI+Science Schmidt Fellows Speaker Series

Organized by the University of Chicago’s Eric and Wendy Schmidt AI in Science Postdoctoral Fellowship Program.

AI+Science Schmidt Fellows Speaker Series: Francisco Villaescusa-Navarro

In this seminar, I will present and describe the CAMELS project, which aims to build bridges between cosmology and galaxy formation by combining numerical simulations and machine learning. Containing a set of more than 10,000 cosmological, both N-body and state-of-the-art hydrodynamic simulations, it is currently the largest dataset of cosmological simulations designed to train artificial intelligence algorithms. I will present some of the results the CAMELS collaboration has obtained recently, such as 1) field-level simulation-based inference while marginalizing over astrophysical effects, 2) discovering fundamental cosmological equations with symbolic regression, 3) weighting the Milky Way and Andromeda with graph neural networks, and 4) the potential of learning cosmology with individual galaxies. I will conclude by presenting a potential strategy the scientific community may pursue to extract the maximum amount of information from cosmological surveys, highlighting the multiple challenges associated with it.

Agenda
4:30pm – 5:15pm: Presentation
5:15pm – 5:30pm: Q&A
5:30pm – 6:00pm: Reception

Meeting location
William Eckhardt Research Center. Room 401
5640 S Ellis Avenue, Chicago, IL 60637
Map It

Francisco Villaescusa-Navarro (Paco) is a research scientist at the Simons Foundation in New York City and a visiting research scholar at Princeton University. He did his Ph.D. at the Instituto de Fisica Corpuscular in Valencia, Spain. Villaescusa-Navarro has been a visiting graduate student at the Canadian Institute for Theoretical Astrophysics (CITA) and the Institute for Theory and Computation (CfA/Harvard University). After completing his Ph.D. he was a postdoctoral researcher at the Astronomical Observatory of Trieste, Italy, and a Flatiron research fellow at the Center for Computational Astrophysics in New York. Later, he went to Princeton University as an associate research scholar.

Villaescusa-Navarro is a theoretical astrophysicist working on several aspects of cosmology. He investigates the properties of the large-scale structure of the Universe with the goal of using cosmological observations to learn about fundamental physics. The goal of his research is to develop the tools that will allow to extract every single bit of cosmological information from cosmic surveys. To achieve this, he combines machine/deep learning techniques with very large sets of state-of-the-art numerical simulations. He is particularly interested in understanding the impact of neutrino masses on cosmological observables, that study combining analytic techniques and state-of-the-art numerical simulations. His research also involves investigating how cosmological and astrophysical information can be extracted from 21cm intensity mapping surveys. Villaescusa-Navarro has also worked on other topics such as the Lyman-alpha forest, modified gravity, BAO reconstruction, kSZ and galaxy formation and evolution. More recently, his research interests have been moving to machine learning, and its usage in cosmology.

Villaescusa-Navarro is the main architect of the Quijote simulations: a suite of 44,110 full N-body simulations containing trillions of particles over a combined cosmological volume larger than the volume of the entire observable Universe. He also developed its precursor, the HADES simulations. He is part of the core team who designed and developed the Cosmology and Astrophysics with MachinE Learning Simulations, CAMELS, a suite of thousands of state-of-the-art cosmological (magneto-)hydrodynamic simulations designed to train machine learning algorithms. Villaescusa-Navarro is also the main developer of the Pylains libraries, a set of python, cython and C libraries designed to analyze numerical simulations. He co-leads the CAMELS project, is a member of Euclid consortium, where he co-leads the Machine Learning for Cosmological Simulations WP, and is part of the PFS, WFIRST, Square Kilometre Array, and SMAUG collaborations.

Parking
Campus North Parking
5505 S Ellis Ave
Map It