Show simple item record

dc.contributor.authorRawlins, T
dc.contributor.authorLewis, A
dc.contributor.authorKipouros, T
dc.date.accessioned2017-12-01T00:27:49Z
dc.date.available2017-12-01T00:27:49Z
dc.date.issued2016
dc.identifier.isbn9781509006199
dc.identifier.doi10.1109/IJCNN.2016.7727321
dc.identifier.urihttp://hdl.handle.net/10072/124117
dc.description.abstractAbstract: A potential area of difficulty for Multi-Objective Optimisation of industrial problems is a class of problems where the majority of the objective space violates blackbox constraints. The difficult arises because potential solutions that violate blackbox constraints provide no information beyond their infeasibility. They provide neither meaningful information about their objective values nor about the degree to which the constraint is violated (or even in some cases which constraint is violated). This means that they do not help to find valid solutions (except by elimination) which, in turn, reduces the early stages of optimisation to effective guesswork until some feasible solutions are found. In this work, we attempt to reduce this problem by using a Decision Tree to identify and repair infeasible solutions by learning the underlying constraints on each parameter. We propose three potential Pre-Repair Methods and compare them on a modified case study of an airfoil lift/drag optimisation problem. Note that no optimisation was done; instead the goal was to decide if the repair methodologies were suitable in the problem space. We used two baselines: not using a Decision Tree, and only using a Decision Tree to identify potentially infeasible solutions for complete regeneration. All three of our proposed methods outperformed the baselines at a statistically significant level of confidence of 0.001.
dc.description.peerreviewedYes
dc.languageEnglish
dc.publisherInstitute of Electrical and Electronics Engineers (IEEE)
dc.publisher.placeUnited States
dc.relation.ispartofconferencenameIJCNN 2016
dc.relation.ispartofconferencetitleProceedings of the International Joint Conference on Neural Networks
dc.relation.ispartofdatefrom2016-07-24
dc.relation.ispartofdateto2016-07-29
dc.relation.ispartoflocationVancouver, Canada
dc.relation.ispartofpagefrom1104
dc.relation.ispartofpageto1111
dc.relation.ispartofvolume2016-October
dc.subject.fieldofresearchNeural, Evolutionary and Fuzzy Computation
dc.subject.fieldofresearchcode080108
dc.titleRepairing Blackbox Constraint Violations in Multi-Objective Optimisation by Use of Decision Trees
dc.typeConference output
dc.type.descriptionE1 - Conferences
dc.type.codeE - Conference Publications
gro.facultyGriffith Sciences, School of Information and Communication Technology
gro.hasfulltextNo Full Text
gro.griffith.authorLewis, Andrew J.
gro.griffith.authorRawlins, Tim


Files in this item

FilesSizeFormatView

There are no files associated with this item.

This item appears in the following Collection(s)

  • Conference outputs
    Contains papers delivered by Griffith authors at national and international conferences.

Show simple item record