BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:{PF}^2ES: Parallel Feasible Pareto Frontier Entropy Search for Mul
 ti-Objective Bayesian Optimization Under Unknown Constraints - Jixiang Qin
 g\, SUMO lab\, Ghent University\,
DTSTART:20230208T110000Z
DTEND:20230208T123000Z
UID:TALK197047@talks.cam.ac.uk
CONTACT:James Allingham
DESCRIPTION:We present Parallel Feasible Pareto Frontier Entropy Search --
 - a novel information-theoretic acquisition function for multi-objective B
 ayesian optimization supporting unknown constraints and batch query. Due t
 o the complexity of characterizing the mutual information between candidat
 e evaluations and (feasible) Pareto frontiers\, existing approaches must e
 ither employ crude approximations that significantly hamper their performa
 nce or rely on expensive inference schemes that substantially increase the
  optimization’s computational overhead. By instead using a variational l
 ower bound\, PF2ES provides a low-cost and accurate estimate of the mutual
  information. Moreover\, we are able to interpret our proposed acquisition
  function by exploring direct links with other popular multi-objective acq
 uisition functions. We benchmark PF2ES against other information-theoretic
  acquisition functions\, demonstrating its competitive performance for opt
 imization across synthetic and real-world design problems.
LOCATION:Cambridge University Engineering Department\, CBL Seminar room BE
 4-38.
END:VEVENT
END:VCALENDAR