Examples of inconsistency in optimization by expected improvement

Research output: Contribution to journalArticlepeer-review

6 Citations (Scopus)

Abstract

We consider the 1D Expected Improvement optimization based on Gaussian processes having spectral densities converging to zero faster than exponentially. We give examples of problems where the optimization trajectory is not dense in the design space. In particular, we prove that for Gaussian kernels there exist smooth objective functions for which the optimization does not converge on the optimum.

Original languageEnglish
Pages (from-to)1773-1790
Number of pages18
JournalJournal of Global Optimization
Volume56
Issue number4
DOIs
Publication statusPublished - Aug 2013
Externally publishedYes

Keywords

  • Analytic kernel
  • Bayesian optimization
  • Expected improvement
  • Gaussian covariance function

Fingerprint

Dive into the research topics of 'Examples of inconsistency in optimization by expected improvement'. Together they form a unique fingerprint.

Cite this