BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Optimization with expensive and uncertain data – challenges and 
 improvements - Coralia Cartis (University of Oxford)
DTSTART:20190502T112000Z
DTEND:20190502T120500Z
UID:TALK123898@talks.cam.ac.uk
CONTACT:Ferdia Sherry
DESCRIPTION:Real-life applications often require the optimization of nonli
 near functions with several unknowns or parameters – where the function 
 is the result of highly expensive and complex model simulations involving 
 noisy data (such as climate or financial models\, chemical experiments)\, 
 or the output of a black-box or legacy code\, that prevent the numerical a
 nalyst from looking inside to find out or calculate problem information su
 ch as derivatives. Thus classical optimization algorithms\, that use deriv
 atives (steepest descent\, Newton’s methods) often fail or are entirely 
 inapplicable in this context. Efficient derivative-free optimization algor
 ithms have been developed in the last 15 years in response to these impera
 tive practical requirements. As even approximate derivatives may be unavai
 lable\, these methods must explore the landscape differently and more crea
 tively. In state of the art techniques\, clouds of points are generated ju
 diciously and sporadically updated to capture local geometries as inexpens
 ively as possible\; local function models around these points are built us
 ing techniques from approximation theory and carefully optimised over a lo
 cal neighbourhood (a trust region) to give a better solution estimate.\n\n
 In this talk\, I will describe our implementations and improvements to sta
 te-of-the-art methods. In the context of the ubiquitous data fitting/least
 -squares applications\, we have developed a simplified approach that is as
  efficient as state of the art in terms of budget use\, while achieving be
 tter scalability. Furthermore\, we substantially improved the robustness o
 f derivative-free methods in the presence of noisy evaluations. Theoretica
 l guarantees of these methods will also be provided. Finally\, despite der
 ivative-free optimisation methods being able to only provably find local o
 ptima\, we illustrate that\, due to their construction and applicability\,
  these methods can offer a practical alternative to global optimisation so
 lvers\, with improved scalability and flexibility. This work is joint with
  Lindon Roberts (Oxford)\, Katya Scheinberg (Lehigh)\, Jan Fiala (NAG Ltd)
  and Benjamin Marteau (NAG Ltd).\n
LOCATION:Centre for Mathematical Sciences\, MR2
END:VEVENT
END:VCALENDAR