BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//euroscipy-2026//speaker//8MTSZC
BEGIN:VTIMEZONE
TZID:Poland
BEGIN:DAYLIGHT
DTSTART:20250721T000000
TZNAME:CEST
TZOFFSETFROM:+0200
TZOFFSETTO:+0200
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20251026T030000
RDATE:20261025T030000
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20260329T030000
RDATE:20270328T030000
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
SUMMARY:setu: Bridging Simulators to Probabilistic Programming in JAX - Ja
 n Boelts (Teusen)\, sethaxen
DTSTART;TZID=Poland:20260721T113000
DTEND;TZID=Poland:20260721T120000
DTSTAMP:20260915T033516Z
UID:pretalx-euroscipy-2026-QJRACH@pretalx.com
DESCRIPTION:Many scientific models\, from climate systems to neural circui
 ts\, are defined as simulators: computer programs that generate data from 
 parameters but provide no tractable likelihood function. This makes them i
 nvisible to probabilistic programming languages (PPLs) like PyMC and NumPy
 ro\, which require explicit likelihoods for Bayesian inference. Practition
 ers are forced to choose: make simplifying assumptions about the simulator
  to use a PPL\, or use the real simulator and give up the rich modeling ca
 pabilities PPLs offer\, such as prior specification\, uncertainty quantifi
 cation and exploitation of hierarchical structures.\n\nWe present `setu` (
 "bridge")\, a JAX-native Python package that closes this gap. `setu` uses 
 generative neural networks trained on simulated data to learn a neural sur
 rogate of the likelihood. This learned likelihood can then be exported dir
 ectly into PPLs via a simple API: `nle.to_pymc()` or `nle.to_numpyro()`. O
 nce inside a PPL\, the full Bayesian toolbox becomes available: hierarchic
 al models\, custom priors\, posterior predictive checks\, and standard MCM
 C samplers — all running on a simulator that was previously out of reach
 .\n\nThe package follows a clean simulate\, train\, validate\, export work
 flow\, with built-in diagnostics to ensure the learned likelihood is trust
 worthy before it ever enters a PPL. In this talk\, we walk through the mot
 ivation\, design\, and a real-world example showing how a black-box simula
 tor gains full PPL capabilities.
LOCATION:Room 2.41 (First Floor\, Turing)
URL:https://pretalx.com/euroscipy-2026/talk/QJRACH/
END:VEVENT
END:VCALENDAR