Show simple item record

dc.contributor.authorDe Ath, G
dc.contributor.authorFieldsend, JE
dc.contributor.authorEverson, RM
dc.date.accessioned2020-05-11T12:26:39Z
dc.date.issued2020-07-08
dc.description.abstractBayesian optimisation is a popular approach for optimising expensive black-box functions. The next location to be evaluated is selected via maximising an acquisition function that balances exploitation and exploration. Gaussian processes, the surrogate models of choice in Bayesian optimisation, are often used with a constant prior mean function equal to the arithmetic mean of the observed function values. We show that the rate of convergence can depend sensitively on the choice of mean function. We empirically investigate 8 mean functions (constant functions equal to the arithmetic mean, minimum, median and maximum of the observed function evaluations, linear, quadratic polynomials, random forests and RBF networks), using 10 synthetic test problems and two real-world problems, and using the Expected Improvement and Upper Confidence Bound acquisition functions. We find that for design dimensions ≥5 using a constant mean function equal to the worst observed quality value is consistently the best choice on the synthetic problems considered. We argue that this worst-observed-quality function promotes exploitation leading to more rapid convergence. However, for the real-world tasks the more complex mean functions capable of modelling the fitness landscape may be effective, although there is no clearly optimum optimum choice.en_GB
dc.description.sponsorshipInnovate UKen_GB
dc.identifier.citationGenetic and Evolutionary Computation Conference Companion (GECCO ’20), 8th July - 12th July 2020. Electronic Only Conference.en_GB
dc.identifier.doi10.1145/3377929.3398118
dc.identifier.grantnumber104400en_GB
dc.identifier.urihttp://hdl.handle.net/10871/120998
dc.language.isoenen_GB
dc.publisherACMen_GB
dc.rights.embargoreasonUnder embargo until conference publication on 8th July 2020.en_GB
dc.rights© 2020 Copyright held by the owner/author(s). Publication rights licensed to ACM.en_GB
dc.subjectBayesian optimisationen_GB
dc.subjectSurrogate modellingen_GB
dc.subjectGaussian processen_GB
dc.subjectMean functionen_GB
dc.subjectacquisition functionen_GB
dc.titleWhat do you mean? The role of the mean function in Bayesian optimisation.en_GB
dc.typeConference proceedingsen_GB
dc.date.available2020-05-11T12:26:39Z
dc.descriptionThis is the author accepted manuscripten_GB
dc.rights.urihttp://www.rioxx.net/licenses/all-rights-reserveden_GB
pubs.funder-ackownledgementYesen_GB
dcterms.dateAccepted2020-05-01
exeter.funder::Innovate UKen_GB
rioxxterms.versionAMen_GB
rioxxterms.licenseref.startdate2020-07-08
rioxxterms.typeConference Paper/Proceeding/Abstracten_GB
refterms.dateFCD2020-05-10T07:13:25Z
refterms.versionFCDAM
refterms.panelBen_GB


Files in this item

This item appears in the following Collection(s)

Show simple item record