Expected improvement criteria for estimating pre-specified features of expensive computer models

Add to your list(s) Download to your calendar using vCal

• Bingham, D (Simon Fraser)
• Wednesday 31 August 2011, 09:30-10:00
• Seminar Room 1, Newton Institute.

If you have a question about this talk, please contact Mustapha Amrani.

This talk has been canceled/deleted

Computer simulators often are inexpensive way to gain insight into a complex physical system, however, it can still be computationally costly. Recent work in this area has focused on experiments where the goal is to fit a response surface or to estimate a particular feature of the surface. In this paper, we focus on the sampling strategies based on a specific criterion called the expected improvement, which often requires fewest number of trials to get a good approximation of the feature of interest. In this paper, we present an overview of the existing expected improvement criteria for some of the common features of interest (e.g., global optimum, contour) in computer experiment setup. We also develop new expected improvement criteria for several process features (e.g., local optima, multiple contours) that are of interest to practitioners.

This talk is part of the Isaac Newton Institute Seminar Series series.

This talk is included in these lists:

This talk is not included in any other list

Note that ex-directory lists are not shown.