BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
PRODID:UW-Madison-Physics-Events
BEGIN:VEVENT
SEQUENCE:0
UID:UW-Physics-Event-7793
DTSTART:20220713T160000Z
DTEND:20220713T171500Z
DTSTAMP:20260414T113519Z
LAST-MODIFIED:20220712T130320Z
LOCATION:Online Seminar: Please sign up for our mailing list at www.ph
 ysicsmeetsml.org for zoom link
SUMMARY:Explainable deep learning models for cosmological structure fo
 rmation\, Physics ∩ ML Seminar\, Luisa Lucie-Smith\, Max Planck Inst
 itute for Astrophysics
DESCRIPTION:According to our standard cosmological model\, the formati
 on of cosmic structures in the Universe is driven by the gravitational
  collapse of small matter density fluctuations present in the early Un
 iverse. The non-linear nature of gravitational collapse makes it diffi
 cult to develop a physical understanding of how complex late-time cosm
 ic structures emerge from these linear initial conditions. In this tal
 k\, I will present an explainable deep learning framework for extracti
 ng new knowledge about the underlying physics of cosmological structur
 e formation. I will focus on an application to dark matter halos\, whi
 ch form the building blocks of cosmic large-scale structure and wherei
 n galaxy formation takes place. The goal is to use interpretable neura
 l networks to discover the independent degrees of freedom in the densi
 ty profiles of dark matter halos. I will show that the model is able t
 o reproduce the known variations encapsulated by previous empirical ap
 proaches. The network then goes further and discovers an additional fa
 ctor of variation in the outer profile\, which we identify as related 
 to infalling dark matter onto the halo (also known as the ‘splashbac
 k’ effect).
URL:https://www.physics.wisc.edu/events/?id=7793
END:VEVENT
END:VCALENDAR