Högskolan i Skövde

his.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • apa-cv
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
A Modular Knowledge-Driven Mutation Operator for Reference-Point Based Evolutionary Algorithms
University of Skövde, School of Engineering Science. University of Skövde, Virtual Engineering Research Environment. (Produktion och Automatiseringsteknik, Production and Automation Engineering)ORCID iD: 0000-0003-3124-0077
University of Skövde, School of Engineering Science. University of Skövde, Virtual Engineering Research Environment. (Produktion och Automatiseringsteknik, Production and Automation Engineering)ORCID iD: 0000-0001-5436-2128
2022 (English)In: IEEE Congress of Evolutionary Computation, CEC - Conference Proceedings, IEEE, 2022Conference paper, Published paper (Refereed)
Abstract [en]

Although an entire frontier of Pareto-optimal solutions exists for multi-objective optimization problems, in practice, decision makers are often only interested in a small subset of these solutions, called the region of interest. Specialized optimizers, such as reference-point based evolutionary algorithms, exist that can focus the search to only find solutions inside this region of interest. These algorithms typically only modify the selection mechanism of regular multi-objective optimizers to preferentially select solutions that conform to the reference point. However, a more effective search may be performed by additionally modifying the variation mechanism of the optimizers, namely the crossover and the mutation operators, to preferentially generate solutions conforming to the reference point. In this paper, we propose a modular mutation operator that uses a recent knowledge discovery technique to first find decision rules unique to the preferred solutions in each generation. These rules are then used to build an empirical distribution in the decision space that can be sampled to generate new mutated solutions which are more likely to be closer to the preferred solutions. The operator is modular in the sense that it can be used with any existing reference-point based evolutionary algorithm by simply replacing the mutation operator. We incorporate the proposed knowledge-driven mutation operator into three such algorithms, and through benchmark test problems up to 10 objectives, demonstrate that their performance improves significantly in the majority of cases according to two different performance indicators. 

Place, publisher, year, edition, pages
IEEE, 2022.
Keywords [en]
Benchmarking, Decision making, Genetic algorithms, Image segmentation, Pareto principle, Modular knowledge, Modulars, Multi-objectives optimization, Mutation operators, Optimizers, Point-based, Preferred solutions, Reference points, Region-of-interest, Regions of interest, Multiobjective optimization, knowledge discovery, multi-objective optimization, mutation operator, reference point
National Category
Computer Sciences Computational Mathematics Other Computer and Information Science
Research subject
Production and Automation Engineering; VF-KDO
Identifiers
URN: urn:nbn:se:his:diva-21917DOI: 10.1109/CEC55065.2022.9870268ISI: 000859282000053Scopus ID: 2-s2.0-85138691252ISBN: 978-1-6654-6708-7 (electronic)ISBN: 978-1-6654-6709-4 (print)OAI: oai:DiVA.org:his-21917DiVA, id: diva2:1701531
Conference
2022 IEEE Congress on Evolutionary Computation, CEC 2022 - Conference Proceedings CEC 2022 Padua 18 July 2022 through 23 July 2022
Part of project
Virtual factories with knowledge-driven optimization (VF-KDO), Knowledge Foundation
Note

© 2022 IEEE.

Available from: 2022-10-06 Created: 2022-10-06 Last updated: 2023-09-01Bibliographically approved
In thesis
1. Knowledge discovery for interactive decision support and knowledge-driven optimization
Open this publication in new window or tab >>Knowledge discovery for interactive decision support and knowledge-driven optimization
2023 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

Multi-objective optimization involves the simultaneous optimization of several objective functions. In real-world problems, these objectives are often in conflict, giving rise to trade-offs in the optimal solutions from the optimization process. All these solutions are equally viable, with no single solution being better or worse than the others. Typically, decision makers have certain preferences that guide the selection of a final solution for practical implementation. While most multi-criteria decision analysis methods focus on the performance of solutions in the objective space, it is important to note that practically relevant knowledge is often found in the design space. Access to this knowledge can provide decision makers with meaningful insights into the problem and the optimization process, leading to more informed decision-making.

This thesis develops and employs methods for knowledge discovery in the context of multi-objective optimization. By emphasizing explicit knowledge representations, this thesis investigates how extracted knowledge can be processed and presented to decision-makers in an interactive manner for insightful decision support. This thesis also explores how extracted knowledge from preferred solutions can be integrated into the algorithms or the multi-objective optimization problem itself, to improve the convergence behavior of optimization algorithms. This approach, called Knowledge-Driven Optimization (KDO), can be implemented either offline or online. Offline KDO involves incorporating knowledge obtained from previous optimization runs into future problem scenarios of a similar nature, restricting the search process to preferred regions of the objective space. A main challenge with such approaches is the storage and retrieval of relevant past knowledge, as well as modifications to the optimization problem formulation. In contrast, online KDO involves integrating knowledge discovery methods with optimization algorithms and utilizing the knowledge obtained during their runtime to enhance the search process, driving algorithms towards better convergence in preferred regions of the objective space. This approach necessitates the development of new search operators capable of incorporating and exploiting various forms of knowledge.

In both offline and online KDO, the veracity and accuracy of the extracted knowledge are critical factors. The thesis validates the effectiveness of the developed methods using various benchmark and engineering optimization test problems, and use-cases from the manufacturing industry. A particular focus is given to generating explicit knowledge that is both meaningful to human decision makers, and can easily be processed algorithmically. The main contributions of this thesis are methods for discovering relevant knowledge about the convergence characteristics of problems, a decision support system for interactive knowledge discovery, and algorithms for realizing both offline and online KDO by incorporating knowledge into the optimization process.

Abstract [sv]

Flermålsoptimering hanterar samtidig optimering av flera målfunktioner, vilka i praktiska optimeringsproblem ofta är i konflikt, vilket ger upphov till avvägningar i de optimala lösningarna från optimeringsprocessen. Alla dessa lösningar är lika värdefulla, och ingen lösning är bättre eller sämre än någon annan. Typiskt sett har beslutsfattare också preferenser som styr valet av en slutlig lösning att implementera i praktiken. De flesta metoder för analys av flera kriterier fokuserar på prestandan hos en uppsättning lösningar i målrymden, det är dock viktigt att notera att praktiskt relevant kunskap ofta finns i designrymden till lösningarna. Tillgång till denna kunskap kan ge beslutsfattare betydelsefulla insikter till både problemet och optimeringsprocessen, vilket leder till mer informerat beslutstagande.

Denna avhandling utvecklar och använder metoder för kunskapsutvinning i sammanhanget av flermålsoptimering. Genom ett särskilt fokus på explicit kunskap, undersöker denna avhandling hur utvunnen kunskap kan bearbetas och presenteras för beslutsfattare på ett interaktivt sätt för förbättrat beslutsstöd. Det undersöks också hur utvunnen kunskap från tidigare lösningar kan integreras i algoritmer för flermålsoptimerings eller direkt i optimeringsproblem för att avlasta beräkning av nya lösningar i optimeringsprocessen. Sådana metoder, kallade kunskapsdriven optimering (KDO), kan implementeras antingen offline eller online. Offline KDO innebär att integrera kunskap som erhållits från tidigare optimeringar, i framtida, liknande problem, vilket avlastar sökprocessen till preferensrika regioner i målrymden. En huvudsaklig utmaning med offline KDO är lagring och återhämtning av relevant tidigare kunskap, samt modifieringar av formuleringar till optimeringsproblem. I kontrast innefattar online KDO att integrera metoder för kunskapsutvinning tillsammans med optimeringsalgoritmer, och att utnyttja den resulterande kunskapen under optimeringen, för att förbättra sökprocessen och driva algoritmerna mot snabbare ankomst i preferensrika regioner i målrymden. Sådana metoder kräver utveckling av nya sökoperatorer kapabla att integrera och utnyttja olika former av utvunnen kunskap.

I både offline och online KDO är det viktigt att den integrerade kunskapen beskriver beslutfattarens preferenser noggrant. Denna avhandling validerar effektiviteten hos de utvecklade metoderna med hjälp av olika benchmark-optimeringsproblem, praktiska tekniska testproblem och fallstudier från tillverkningsindustrin. Ett särskilt fokus har lagts på utvinning av explicit kunskap som både är meningsfull för beslutsfattare och som enkelt kan bearbetas algoritmiskt. Denna avhandlings huvudsakliga bidrag består av metoder för utvinning av relevant kunskap om sökbeteendet för problem, ett beslutstödssystem för interaktiv kunskapsutvinning, samt algoritmer för att förverkliga både offline och online KDO genom att integrera kunskap i optimeringsprocessen.

Place, publisher, year, edition, pages
Skövde: University of Skövde, 2023. p. xv, 185
Series
Dissertation Series ; 52
National Category
Computer Sciences Information Systems Software Engineering Computer Systems Computational Mathematics
Research subject
Virtual Production Development (VPD); VF-KDO
Identifiers
urn:nbn:se:his:diva-23152 (URN)978-91-987906-6-5 (ISBN)
Public defence
2023-09-27, Assar Industrial Innovation Arena, Kavelbrovägen 2B, Skövde, 13:00 (English)
Opponent
Supervisors
Available from: 2023-09-01 Created: 2023-09-01 Last updated: 2023-09-01Bibliographically approved

Open Access in DiVA

No full text in DiVA

Other links

Publisher's full textScopus

Authority records

Smedberg, HenrikBandaru, Sunith

Search in DiVA

By author/editor
Smedberg, HenrikBandaru, Sunith
By organisation
School of Engineering ScienceVirtual Engineering Research Environment
Computer SciencesComputational MathematicsOther Computer and Information Science

Search outside of DiVA

GoogleGoogle Scholar

doi
isbn
urn-nbn

Altmetric score

doi
isbn
urn-nbn
Total: 132 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • apa-cv
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf