BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Bayes Reading Group:  Hugh Dance (UCL)
DTSTART:20251219T160000Z
DTEND:20251219T170000Z
DTSTAMP:20261009T021900Z
UID:indico-event-4111@indico.flatironinstitute.org
DESCRIPTION:Discussion lead:  Hugh Dance (UCL)\nTitle: Efficiently Vec
 torized MCMC on Modern Accelerators\nAbstract: With the advent of automat
 ic vectorization tools  writing multi-chain MCMC algorithms is often now 
 as simple as invoking those tools on single-chain code. Whilst convenient\
 , for various MCMC algorithms this results in a synchronization problem --
  loosely speaking\, at each iteration all chains running in parallel must 
 wait until the last chain has finished drawing its sample. In this work\, 
 we show how to design single-chain MCMC algorithms in a way that avoids sy
 nchronization overheads when deploying automatic vectorization tools\, via
  the framework of finite state machines (FSMs). Using a simplified model\,
  we derive an exact theoretical form of the obtainable speed-ups using our
  approach\, and use it to make principled recommendations for optimal algo
 rithm design. We implement several popular MCMC algorithms as FSMs\, inclu
 ding Elliptical Slice Sampling\, HMC-NUTS\, and Delayed Rejection\, demons
 trating speed-ups of up to an order of magnitude in experiment\n\nhttps://
 indico.flatironinstitute.org/event/4111/
LOCATION:3rd Floor Conference Room  (162 Fifth Avenue )
URL:https://indico.flatironinstitute.org/event/4111/
END:VEVENT
END:VCALENDAR