Journal article 918 views 117 downloads
Predicting effective control parameters for differential evolution using cluster analysis of objective function features
Journal of Heuristics, Volume: 25, Issue: 6, Pages: 1015 - 1031
Swansea University Authors: Sean Walton , Rowan Brown
-
PDF | Version of Record
Released under the terms of a Creative Commons Attribution 4.0 International License (CC-BY).
Download (1.21MB)
DOI (Published version): 10.1007/s10732-019-09419-8
Abstract
A methodology is introduced which uses three simple objectivefunction features to predict effective control parameters for differential evolution. This is achieved using cluster analysis techniques to classify objectivefunctions using these features. Information on prior performance of variouscontro...
Published in: | Journal of Heuristics |
---|---|
ISSN: | 1381-1231 1572-9397 |
Published: |
Springer Science and Business Media LLC
2019
|
Online Access: |
Check full text
|
URI: | https://cronfa.swan.ac.uk/Record/cronfa50868 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
first_indexed |
2019-06-17T20:52:36Z |
---|---|
last_indexed |
2020-07-31T19:12:12Z |
id |
cronfa50868 |
recordtype |
SURis |
fullrecord |
<?xml version="1.0"?><rfc1807><datestamp>2020-07-31T14:31:26.9701473</datestamp><bib-version>v2</bib-version><id>50868</id><entry>2019-06-17</entry><title>Predicting effective control parameters for differential evolution using cluster analysis of objective function features</title><swanseaauthors><author><sid>0ec10d5e3ed3720a2d578417a894cf49</sid><ORCID>0000-0002-6451-265X</ORCID><firstname>Sean</firstname><surname>Walton</surname><name>Sean Walton</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>d7db8d42c476dfa69c15ce06d29bd863</sid><ORCID>0000-0003-3628-2524</ORCID><firstname>Rowan</firstname><surname>Brown</surname><name>Rowan Brown</name><active>true</active><ethesisStudent>false</ethesisStudent></author></swanseaauthors><date>2019-06-17</date><deptcode>SCS</deptcode><abstract>A methodology is introduced which uses three simple objectivefunction features to predict effective control parameters for differential evolution. This is achieved using cluster analysis techniques to classify objectivefunctions using these features. Information on prior performance of variouscontrol parameters for each classification is then used to determine which control parameters to use in future optimisations. Our approach is compared tostate–of–the–art adaptive and non–adaptive techniques. Two accepted benchmark suites are used to compare performance and in all cases we show thatthe improvement resulting from our approach is statistically significant. Themajority of the computational effort of this methodology is performed off–line, however even when taking into account the additional on–line cost ourapproach outperforms other adaptive techniques. We also study the key tuning parameters of our methodology, such as number of clusters, which furthersupport the finding that the simple features selected are predictors of effectivecontrol parameters. The findings presented in this paper are significant becausethey show that simple to calculate features of objective functions can help toselect control parameters for optimisation algorithms. This can have an immediate positive impact the application of these optimisation algorithms on realworld problems where it is often difficult to select effective control parameters.</abstract><type>Journal Article</type><journal>Journal of Heuristics</journal><volume>25</volume><journalNumber>6</journalNumber><paginationStart>1015</paginationStart><paginationEnd>1031</paginationEnd><publisher>Springer Science and Business Media LLC</publisher><issnPrint>1381-1231</issnPrint><issnElectronic>1572-9397</issnElectronic><keywords>tuning evolutionary algorithms, differential evolution, unsupervised learning, optimization methods, evolutionary algorithms</keywords><publishedDay>1</publishedDay><publishedMonth>12</publishedMonth><publishedYear>2019</publishedYear><publishedDate>2019-12-01</publishedDate><doi>10.1007/s10732-019-09419-8</doi><url>http://dx.doi.org/10.1007/s10732-019-09419-8</url><notes>Originality: The first results showing that a statistically significant improvement in performance can be achieved for optimisation algorithms using automatic problem classification.Rigour: The paper went through a lengthy and rigorous peer review process because our results were so significant. A total of 67,800 experiments were run over a variety of established benchmark functions, repeated for different random seeds. Significance: Our results show that you can use information about a function to predict how to tune your optimisation algorithm for it. This is a step towards fixing a key limitation of many gradient free optimisation algorithms – problem dependent tuning.</notes><college>COLLEGE NANME</college><department>Computer Science</department><CollegeCode>COLLEGE CODE</CollegeCode><DepartmentCode>SCS</DepartmentCode><institution>Swansea University</institution><apcterm/><lastEdited>2020-07-31T14:31:26.9701473</lastEdited><Created>2019-06-17T15:15:33.1182556</Created><authors><author><firstname>Sean</firstname><surname>Walton</surname><orcid>0000-0002-6451-265X</orcid><order>1</order></author><author><firstname>Rowan</firstname><surname>Brown</surname><orcid>0000-0003-3628-2524</orcid><order>2</order></author></authors><documents><document><filename>0050868-09072019160312.pdf</filename><originalFilename>50868.pdf</originalFilename><uploaded>2019-07-09T16:03:12.3530000</uploaded><type>Output</type><contentLength>1344144</contentLength><contentType>application/pdf</contentType><version>Version of Record</version><cronfaStatus>true</cronfaStatus><embargoDate>2019-07-08T00:00:00.0000000</embargoDate><documentNotes>Released under the terms of a Creative Commons Attribution 4.0 International License (CC-BY).</documentNotes><copyrightCorrect>true</copyrightCorrect><language>eng</language></document></documents><OutputDurs/></rfc1807> |
spelling |
2020-07-31T14:31:26.9701473 v2 50868 2019-06-17 Predicting effective control parameters for differential evolution using cluster analysis of objective function features 0ec10d5e3ed3720a2d578417a894cf49 0000-0002-6451-265X Sean Walton Sean Walton true false d7db8d42c476dfa69c15ce06d29bd863 0000-0003-3628-2524 Rowan Brown Rowan Brown true false 2019-06-17 SCS A methodology is introduced which uses three simple objectivefunction features to predict effective control parameters for differential evolution. This is achieved using cluster analysis techniques to classify objectivefunctions using these features. Information on prior performance of variouscontrol parameters for each classification is then used to determine which control parameters to use in future optimisations. Our approach is compared tostate–of–the–art adaptive and non–adaptive techniques. Two accepted benchmark suites are used to compare performance and in all cases we show thatthe improvement resulting from our approach is statistically significant. Themajority of the computational effort of this methodology is performed off–line, however even when taking into account the additional on–line cost ourapproach outperforms other adaptive techniques. We also study the key tuning parameters of our methodology, such as number of clusters, which furthersupport the finding that the simple features selected are predictors of effectivecontrol parameters. The findings presented in this paper are significant becausethey show that simple to calculate features of objective functions can help toselect control parameters for optimisation algorithms. This can have an immediate positive impact the application of these optimisation algorithms on realworld problems where it is often difficult to select effective control parameters. Journal Article Journal of Heuristics 25 6 1015 1031 Springer Science and Business Media LLC 1381-1231 1572-9397 tuning evolutionary algorithms, differential evolution, unsupervised learning, optimization methods, evolutionary algorithms 1 12 2019 2019-12-01 10.1007/s10732-019-09419-8 http://dx.doi.org/10.1007/s10732-019-09419-8 Originality: The first results showing that a statistically significant improvement in performance can be achieved for optimisation algorithms using automatic problem classification.Rigour: The paper went through a lengthy and rigorous peer review process because our results were so significant. A total of 67,800 experiments were run over a variety of established benchmark functions, repeated for different random seeds. Significance: Our results show that you can use information about a function to predict how to tune your optimisation algorithm for it. This is a step towards fixing a key limitation of many gradient free optimisation algorithms – problem dependent tuning. COLLEGE NANME Computer Science COLLEGE CODE SCS Swansea University 2020-07-31T14:31:26.9701473 2019-06-17T15:15:33.1182556 Sean Walton 0000-0002-6451-265X 1 Rowan Brown 0000-0003-3628-2524 2 0050868-09072019160312.pdf 50868.pdf 2019-07-09T16:03:12.3530000 Output 1344144 application/pdf Version of Record true 2019-07-08T00:00:00.0000000 Released under the terms of a Creative Commons Attribution 4.0 International License (CC-BY). true eng |
title |
Predicting effective control parameters for differential evolution using cluster analysis of objective function features |
spellingShingle |
Predicting effective control parameters for differential evolution using cluster analysis of objective function features Sean Walton Rowan Brown |
title_short |
Predicting effective control parameters for differential evolution using cluster analysis of objective function features |
title_full |
Predicting effective control parameters for differential evolution using cluster analysis of objective function features |
title_fullStr |
Predicting effective control parameters for differential evolution using cluster analysis of objective function features |
title_full_unstemmed |
Predicting effective control parameters for differential evolution using cluster analysis of objective function features |
title_sort |
Predicting effective control parameters for differential evolution using cluster analysis of objective function features |
author_id_str_mv |
0ec10d5e3ed3720a2d578417a894cf49 d7db8d42c476dfa69c15ce06d29bd863 |
author_id_fullname_str_mv |
0ec10d5e3ed3720a2d578417a894cf49_***_Sean Walton d7db8d42c476dfa69c15ce06d29bd863_***_Rowan Brown |
author |
Sean Walton Rowan Brown |
author2 |
Sean Walton Rowan Brown |
format |
Journal article |
container_title |
Journal of Heuristics |
container_volume |
25 |
container_issue |
6 |
container_start_page |
1015 |
publishDate |
2019 |
institution |
Swansea University |
issn |
1381-1231 1572-9397 |
doi_str_mv |
10.1007/s10732-019-09419-8 |
publisher |
Springer Science and Business Media LLC |
url |
http://dx.doi.org/10.1007/s10732-019-09419-8 |
document_store_str |
1 |
active_str |
0 |
description |
A methodology is introduced which uses three simple objectivefunction features to predict effective control parameters for differential evolution. This is achieved using cluster analysis techniques to classify objectivefunctions using these features. Information on prior performance of variouscontrol parameters for each classification is then used to determine which control parameters to use in future optimisations. Our approach is compared tostate–of–the–art adaptive and non–adaptive techniques. Two accepted benchmark suites are used to compare performance and in all cases we show thatthe improvement resulting from our approach is statistically significant. Themajority of the computational effort of this methodology is performed off–line, however even when taking into account the additional on–line cost ourapproach outperforms other adaptive techniques. We also study the key tuning parameters of our methodology, such as number of clusters, which furthersupport the finding that the simple features selected are predictors of effectivecontrol parameters. The findings presented in this paper are significant becausethey show that simple to calculate features of objective functions can help toselect control parameters for optimisation algorithms. This can have an immediate positive impact the application of these optimisation algorithms on realworld problems where it is often difficult to select effective control parameters. |
published_date |
2019-12-01T04:02:32Z |
_version_ |
1763753226039984128 |
score |
11.033112 |