BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Sabre//Sabre VObject 4.5.8//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:Europe/Zurich X-LIC-LOCATION:Europe/Zurich TZURL:http://tzurl.org/zoneinfo/Europe/Zurich BEGIN:DAYLIGHT TZOFFSETFROM:+0100 TZOFFSETTO:+0200 TZNAME:CEST DTSTART:19810329T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET DTSTART:19961027T030000 RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT UID:news1924@dmi.unibas.ch DTSTAMP;TZID=Europe/Zurich:20251022T110232 DTSTART;TZID=Europe/Zurich:20251104T161500 SUMMARY:Simulation-based inference of cosmology using neural compression of multi-probe maps DESCRIPTION:Ongoing galaxy surveys are designed to observe the large-scale structure of the Universe using a number of cosmological probes\, includin g weak gravitational lensing and galaxy clustering. Conventionally\, const raints on the cosmological parameters are obtained by comparing two-point statistics of the observables with semi-analytical theory predictions. How ever\, we know that due to nonlinear structure formation at late times\, t he physical fields contain non-Gaussian information which is not captured at the two-point level.\\r\\nIn this talk\, I introduce a pipeline that le verages numerical theory predictions and the expressive power of deep lear ning to extract this additional cosmological information by learning the ( non-Gaussian) summary statistic from forward modeled maps instead. More sp ecifically\, I present our analysis framework for simulation-based inferen ce in three parts: 1) a forward model to generate over one million survey- realistic weak lensing and galaxy clustering maps from the CosmoGridV1 sim ulation suite\; 2) graph convolutional networks that compress these maps i nto low-dimensional\, maximally informative summary statistics\; and 3) st andard normalizing flows approximating the unknown likelihood to infer pos terior distributions of cosmological parameters\, given a (mock) observati on.\\r\\nArne Thomsen [https://www.phys.ethz.ch/the-department/people/pers on-detail.MjMwMzU1.TGlzdC81MTUsMTE3MjU5OTI5OQ==.html] X-ALT-DESC:

Ongoing galaxy surveys are designed to observe the large-scal e structure of the Universe using a number of cosmological probes\, includ ing weak gravitational lensing and galaxy clustering. Conventionally\, con straints on the cosmological parameters are obtained by comparing two-poin t statistics of the observables with semi-analytical theory predictions. H owever\, we know that due to nonlinear structure formation at late times\, the physical fields contain non-Gaussian information which is not capture d at the two-point level.

\n

In this talk\, I introduce a pipeline th at leverages numerical theory predictions and the expressive power of deep learning to extract this additional cosmological information by learning the (non-Gaussian) summary statistic from forward modeled maps instead. Mo re specifically\, I present our analysis framework for simulation-based in ference in three parts: 1) a forward model to generate over one million su rvey-realistic weak lensing and galaxy clustering maps from the CosmoGridV 1 simulation suite\; 2) graph convolutional networks that compress these m aps into low-dimensional\, maximally informative summary statistics\; and 3) standard normalizing flows approximating the unknown likelihood to infe r posterior distributions of cosmological parameters\, given a (mock) obse rvation.

\n

Arne Thomsen< /p> END:VEVENT END:VCALENDAR