BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Acceleration of MCMC methods by non-reversibility and factorizatio
n
DTSTART;VALUE=DATE-TIME:20190621T123000Z
DTEND;VALUE=DATE-TIME:20190621T143000Z
DTSTAMP;VALUE=DATE-TIME:20190921T051819Z
UID:indico-event-19377@indico.in2p3.fr
DESCRIPTION:During this talk\, I will present non-reversible Markov-chain\
n Monte Carlo methods\, based on piecewise deterministic Markov\n
processes (PDMP). First developed for multiparticle systems\, the
\n goal was to emulate the successes of cluster algorithms for spi
n\n systems and was achieved through the replacement of the time\n
reversibility by symmetries of the sampled probability\n
distribution itself. These methods have shown to bring clear\n acc
elerations and are now competing with molecular dynamics\n methods
in chemical physics or state-of-the-art sampling schemes\,\n e.g.
Hamiltonian Monte Carlo\, in statistical inference. Finally\, I\n
will explain how the factorization of interaction terms can lead\n
to computational complexity reduction\, for instance in presence of\n
long-range interactions.\n\n \n\n \n\n\n\n \n\n \n\nhttps://ind
ico.in2p3.fr/event/19377/
LOCATION:LPC Amphi Recherche
URL:https://indico.in2p3.fr/event/19377/
END:VEVENT
END:VCALENDAR