BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:ABC for galaxy star formation history model choice
DTSTART;VALUE=DATE-TIME:20190920T093000Z
DTEND;VALUE=DATE-TIME:20190920T101500Z
DTSTAMP;VALUE=DATE-TIME:20220816T003106Z
UID:indico-contribution-4790-4040@indico.math.cnrs.fr
DESCRIPTION:Speakers: Grégoire Aufort (Institut de Mathématiques de Mars
eille (I2M) Aix-Marseille Université)\nWe are interested in the bayesian
model choice problem when a large number of objects have to be process
ed. We propose an extension of the ABC-RandomForest algorithm for
model choice\, based on crossentropy minimization on the ABC si
mulation catalogue. This learning algorithm allows us to bypass the use o
f summary statistics for ABC. We present an application in astrophysics.
From photometric data\, we show the relevance of the complexification of
a stellar formation history model for an important part of the datasets am
ong tens of thousands of galaxies.\n\nhttps://indico.math.cnrs.fr/event/47
90/contributions/4040/
LOCATION:Clermont-Ferrand
URL:https://indico.math.cnrs.fr/event/4790/contributions/4040/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Bayesian inference with black-box cosmological models
DTSTART;VALUE=DATE-TIME:20190920T081500Z
DTEND;VALUE=DATE-TIME:20190920T091500Z
DTSTAMP;VALUE=DATE-TIME:20220816T003106Z
UID:indico-contribution-4790-3918@indico.math.cnrs.fr
DESCRIPTION:Speakers: Florent Leclercq (Imperial Centre for Inference and
Cosmology\, London)\nLarge-scale astronomical surveys carry opportunities
for testing physical theories about the origin and evolution of the Univer
se. Advancing the research frontier requires solving challenging and uniqu
e statistical problems\, to unlock the information content of massive and
complex data streams. In this talk\, I will present recent methodological
advances\, aiming at fitting cosmological data with "black-box" numerical
models. I will discuss two different solutions\, depending on the scenario
: Bayesian optimisation (BOLFI) and Taylor-expansion of the simulator (SEL
FI).\n\nReferences:\nhttp://arxiv.org/abs/1805.07152\nhttp://arxiv.org/abs
/1902.10149\n\nhttps://indico.math.cnrs.fr/event/4790/contributions/3918/
LOCATION:Clermont-Ferrand
URL:https://indico.math.cnrs.fr/event/4790/contributions/3918/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Gibbs-ABC
DTSTART;VALUE=DATE-TIME:20190920T071500Z
DTEND;VALUE=DATE-TIME:20190920T081500Z
DTSTAMP;VALUE=DATE-TIME:20220816T003106Z
UID:indico-contribution-4790-3919@indico.math.cnrs.fr
DESCRIPTION:Speakers: Christian Robert (CEREMADE\, Paris)\nApproximate Bay
esian computation methods are useful for generative models with intractabl
e likelihoods. These methods are however sensitive to the dimension of the
parameter space\, requiring exponentially increasing resources as this di
mension grows. To tackle this difficulty\, we explore a Gibbs version of t
he ABC approach that runs component-wise approximate Bayesian computation
steps aimed at the corresponding conditional posterior distributions\, and
based on summary statistics of reduced dimensions. While lacking the stan
dard justifications for the Gibbs sampler\, the resulting Markov chain is
shown to converge in distribution under some partial independence conditio
ns. The associated stationary distribution can further be shown to be clos
e to the true posterior distribution and some hierarchical versions of the
proposed mechanism enjoy a closed form limiting distribution. Experiments
also demonstrate the gain in efficiency brought by the Gibbs version over
the standard solution.\n\nhttps://arxiv.org/abs/1905.13599\n\nhttps://ind
ico.math.cnrs.fr/event/4790/contributions/3919/
LOCATION:Clermont-Ferrand
URL:https://indico.math.cnrs.fr/event/4790/contributions/3919/
END:VEVENT
BEGIN:VEVENT
SUMMARY:ABC in cosmology: Likelihood-free inference without the inverse co
variance matrix
DTSTART;VALUE=DATE-TIME:20190920T121500Z
DTEND;VALUE=DATE-TIME:20190920T131500Z
DTSTAMP;VALUE=DATE-TIME:20220816T003106Z
UID:indico-contribution-4790-3921@indico.math.cnrs.fr
DESCRIPTION:Speakers: Martin Kilbinger (CEA Saclay)\nIn traditional likeli
hood-based parameter inference methods\, the inverse of the data covarianc
e matrix has to be computed. In cosmology\, the covariance is often estima
ted from expensive numerical simulations. Limits on the allowed biases on
parameter constraints from the inversion of the noisy\, high-dimensional c
ovariance matrix sets strong requirements on the necessary number of simul
ations\, which has to be much larger than the data dimension. For a realis
tic setting of typical cosmological data\, this number can be in the thous
ands\, making the use of time-consuming N-body simulations prohibitive.\n\
nIn this talk I propose to use Approximate Bayesian Computation (ABC) as a
likelihood-free inference method to obtain constraints on cosmological pa
rameters. Model simulations of the data vector are obtained quickly by dra
wing from an analytical multi-variate distribution\, requiring only a cova
riance matrix but not its inverse. Using toy models\, I show that the numb
er of simulations can be much smaller than the data dimension. I present f
irst results from applying ABC to weak gravitational lensing\, which is on
e of the main cosmological probes to explore dark energy and the dark-matt
er distribution in the universe.\n\nhttps://indico.math.cnrs.fr/event/4790
/contributions/3921/
LOCATION:Clermont-Ferrand
URL:https://indico.math.cnrs.fr/event/4790/contributions/3921/
END:VEVENT
END:VCALENDAR