dc.contributor.author | Fieldsend, JE | |
dc.contributor.author | Alyahya, K | |
dc.date.accessioned | 2019-04-24T13:09:27Z | |
dc.date.issued | 2019-07-13 | |
dc.description.abstract | Local optima networks (LONs) represent the landscape of optimisation problems. In a LON, graph vertices represent local optima in the search domain, their radii the basin sizes, and directed edges between vertices the ability to transit from one basin to another (with the edge width denoting how easy this is). Recently, a network construction approach inspired by LONs has been proposed for multi-objective problems which uses an undirected graph, representing mutually non-dominating solutions and neighbouring links, but not basin sizes. In contrast, here we introduce two formulations for multi/many-objective problems which are analogous to the traditional LON, using dominance-based hill-climbing to characterise the search domain. Each vertex represents a set of locally optimal solutions, with basins and ease of transition between them shown. These LONs vary depending on whether a point-based (dominance neutral optima) or set-based (Pareto local optima) representation is used to define mode construction. We illustrate these alternative formulations on some illustrative problems.We discuss some of the underlying computational issues in constructing LONs in a multiobjective as opposed to uni-objective problem domain, along with the inherent issue of neutrality — as each a vertex in these graphs almost invariably represents a set in our proposed constructs. | en_GB |
dc.description.sponsorship | Engineering and Physical Sciences Research Council (EPSRC) | en_GB |
dc.identifier.citation | GECCO '19: Genetic and Evolutionary Computation Conference, 13-17 July 2019, Prague, Czech Republic | en_GB |
dc.identifier.doi | 10.1145/3319619.3326838 | |
dc.identifier.grantnumber | EP/N017846/1 | en_GB |
dc.identifier.uri | http://hdl.handle.net/10871/36891 | |
dc.language.iso | en | en_GB |
dc.publisher | Association for Computing Machinery (ACM) | en_GB |
dc.relation.url | https://github.com/fieldsend/mo_lons | |
dc.rights | © 2019 Copyright held by the owner/author(s). Publication rights licensed to ACM | en_GB |
dc.subject | multi-objective | en_GB |
dc.subject | optimisation | en_GB |
dc.subject | fitness landscapes | en_GB |
dc.title | Visualising the Landscape of Multi-Objective Problems using Local Optima Networks | en_GB |
dc.type | Conference proceedings | en_GB |
dc.date.available | 2019-04-24T13:09:27Z | |
dc.identifier.isbn | 978-1-4503-6748-6 | |
dc.description | This is the author accepted manuscript. The final version is available from ACM via the DOI in this record | en_GB |
dc.description | The codebase for this paper is available at https://github.com/fieldsend/mo_lons | |
dc.rights.uri | http://www.rioxx.net/licenses/all-rights-reserved | en_GB |
pubs.funder-ackownledgement | Yes | en_GB |
dcterms.dateAccepted | 2019-04-18 | |
exeter.funder | ::Engineering and Physical Sciences Research Council (EPSRC) | en_GB |
rioxxterms.version | AM | en_GB |
rioxxterms.licenseref.startdate | 2019-04-18 | |
rioxxterms.type | Conference Paper/Proceeding/Abstract | en_GB |
refterms.dateFCD | 2019-04-23T15:33:06Z | |
refterms.versionFCD | AM | |
refterms.dateFOA | 2019-04-29T10:24:46Z | |
refterms.panel | B | en_GB |