Efficient Global Optimization of Expensive Black-Box Functions

  • Authors:

  • Donald R. Jones;Matthias Schonlau;William J. Welch

  • Affiliations:

  • Operations Research Department, General Motors R&/D Operations, Warren, MI, USA/;National Institute of Statistical Sciences, Research Triangle Park, NC, USA/;Department of Statistics and Actuarial Science and The Institute for Improvement in Quality and Productivity, University of Waterloo, Waterloo, Ontario, Canada

  • Venue:
  • Journal of Global Optimization
  • Year:
  • 1998

Quantified Score

Hi-index 0.00

Visualization

Abstract

In many engineering optimization problems, the number of functionevaluations is severely limited by time or cost. These problems pose aspecial challenge to the field of global optimization, since existingmethods often require more function evaluations than can be comfortablyafforded. One way to address this challenge is to fit response surfaces todata collected by evaluating the objective and constraint functions at a fewpoints. These surfaces can then be used for visualization, tradeoffanalysis, and optimization. In this paper, we introduce the reader to aresponse surface methodology that is especially good at modeling thenonlinear, multimodal functions that often occur in engineering. We thenshow how these approximating functions can be used to construct an efficientglobal optimization algorithm with a credible stopping rule. The key tousing response surfaces for global optimization lies in balancing the needto exploit the approximating surface (by sampling where it is minimized)with the need to improve the approximation (by sampling where predictionerror may be high). Striking this balance requires solving certain auxiliaryproblems which have previously been considered intractable, but we show howthese computational obstacles can be overcome.