BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Explainable deep learning models in cosmology - Luisa Lucie-Smith 
 (University of Hamburg)
DTSTART:20250217T130000Z
DTEND:20250217T140000Z
UID:TALK221773@talks.cam.ac.uk
CONTACT:Thomas Colas
DESCRIPTION:Machine learning has significantly improved the way cosmologis
 ts model and interpret cosmological data\; yet\, its "black box" nature of
 ten limits our ability to trust and understand its results. In this talk\,
  I will present an explainable deep learning framework designed to rely on
  a minimal set of physically interpretable parameters which describe the d
 ata. I will first discuss applications to dark matter halos\, demonstratin
 g how these neural networks can be used to model their final properties 
 — such as their density profiles — and connect them to the underlying 
 physics. Additionally\, I will present applications to the cosmic microwav
 e background\, revealing to which parameters the CMB temperature power spe
 ctrum is sensitive in the context of early dark energy models.
LOCATION:CMS\, Pav. B\, CTC Common Room (B1.19) [Potter Room]
END:VEVENT
END:VCALENDAR