BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:MSG Design of Experiments Seminar Series: Mutual Information for C
 omputer Experiments (MICE): design\, optimization\, and data assimilation:
  applications to tsunami hazard - Serge Guillas (University College London
 )
DTSTART:20180620T151000Z
DTEND:20180620T160000Z
UID:TALK107149@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:We present a new method for the design of computer experiments
 . The  sequential design algorithm MICE (Mutual Information for Computer  
 Experiments) adaptively selects the input values at which to run the  comp
 uter simulator\, in order to maximize the expected information gain  (mutu
 al information) over the input space. The superior computational  efficien
 cy of MICE compared to other algorithms is demonstrated on test  functions
 \, and on the tsunami model VOLNA with overall gains of 20-50%.  Moreover\
 , there is a clear computational advantage in building a design  of comput
 er experiments solely on a subset of active variables. However\,  this pri
 or selection inflates the limited computational budget. We thus  interweav
 e MICE with a screening algorithm to improve the overall  efficiency of bu
 ilding an emulator. This approach allows us to assess  future tsunami risk
  for complex earthquake sources over Cascadia. An  application to optimiza
 tion of expensive black-box functions using MICE  is also introduced. It i
 s then employed in a data assimilation scheme to  design an optimal networ
 k of buoys near shore for the purpose of  detecting incoming tsunamis.<br>
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR