Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.486780
Title: Improving Convergence, Diversity and Pertinency in Multiobjective Optimisation
Author: Adra, Salem Fawaz
ISNI:       0000 0001 3398 7161
Awarding Body: University of Sheffield
Current Institution: University of Sheffield
Date of Award: 2007
Availability of Full Text:
Access through EThOS:
Abstract:
Real-world problems commonly require the simultaneous consideration of multiple, often conflicting, objectives. Solving a multiobjective optimisation problem (MOP) is concerned with finding an ideal set of tradeoff solutions which are close to and uniformly distributed across the optimal tradeoff surface. Convergence and diversity are thus essential requirements .of multiobjective optimisers, which are sometimes also required to focus on pertinent areas of the search space. Evolutionary computation (EC) techniques are stochastic, population-based, global search techniques well suited for solving MOPs. However, EC techniques can often involve a large number of objective function calculations which can make the convergence towards optimal tradeoff surfaces computationally expensive. Additionally, in the evolutionary multiobjective optimization community, the bi-objective case is the most heavily studied. Conclusions drawn from such low-dimensional frameworks used to be generalized for all problems' dimensions. Research, however, has shown that high-dimensional problems (> 3 objectives) can possess different characteristics. One of the most important challenges faced in such optimisation scenarios is the conflict between convergence and diversity of solutions. In this study, new approaches are proposed for enhancing the convergence and diversification capacities of so~e of the best multiobjective evolutionary optimisers (MOEAs). The inclusion of quality metrlcs as indicators is implemented as an approach for solving the conflict between solutions' convergence and diversity in high-dimensional optimisation problems. Moreover, a convergence acceleration technique for MOEAs which exploits the objective space, where the goal and objectives lies,is devised and assessed. In the final part of the study, some established progressive preference articulation techniques are examined, and their utility for tackling MOPs is discussed from the viewpoint of the decision maker. Progressive preference articulation techniques are effective methods for supporting the decision maker in guiding the search into pertinent regions of interest and coping with the curse of dimensionality.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.486780  DOI: Not available
Share: