BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Theoretical Physics Colloquium
SUMMARY:Comparing gravitational-wave data and stellar-phys
 ics predictions with deep learning - Prof. Davide 
 Gerosa\, University of Milano-Bicocca
DTSTART;TZID=Europe/London:20230215T140000
DTEND;TZID=Europe/London:20230215T150000
UID:TALK197269AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/197269
DESCRIPTION:The catalog of gravitational-wave events is growin
 g\, and so are our hopes of constraining the under
 lying astrophysics of stellar-mass black-hole merg
 ers by inferring the distributions of\, e.g.\, mas
 ses and spins. While conventional analyses paramet
 rize this population with simple phenomenological 
 models\, we propose an innovative physics-first ap
 proach that compares gravitational-wave data again
 st astrophysical simulations. We combine state-of-
 the-art deep-learning techniques with hierarchical
  Bayesian inference and exploit our approach to co
 nstrain the properties of repeated black-hole merg
 ers from the gravitational-wave events in the most
  recent LIGO/Virgo catalog. Deep neural networks a
 llow us to (i) construct a flexible population mod
 el that accurately emulates simulations of hierarc
 hical mergers\, (ii) estimate selection effects\, 
 and (iii) recover the branching ratios of repeated
 -merger generations. Among our results we find tha
 t: the distribution of host-environment escape spe
 eds favors values <100 km/s but is relatively flat
 \; first-generation black holes are born with a ma
 ximum mass that is compatible with current estimat
 es from pair-instability supernovae\; there is mul
 timodal substructure in both the mass and spin dis
 tributions due to repeated mergers\; and binaries 
 with a higher-generation component make up at leas
 t 15% of the underlying population. The deep-learn
 ing pipeline we present is ready to be used in con
 junction with realistic astrophysical population-s
 ynthesis predictions.
LOCATION:Both in-person (at MR2\, DAMTP) and online (detail
 s to be sent by email)
CONTACT:Isobel Romero-Shaw
END:VEVENT
END:VCALENDAR