In evolutionary optimization, high dimensionality has a negative impact on the algorithms performance since the complexity of handled problems grows exponentially with dimensionality. Cooperative co-evolutionary algorithms overstep this limitation by enjoying the benefits of the divide-and-conquer strategy. They divide large problems into some simpler and smaller subproblems that can be optimized with a standard evolutionary algorithm. The aim of this thesis is to study these cooperative co-evolutionary algorithms and to develop new tools to allow them to tackle optimization problems with common features appearing in engineering and sciences. Particular attention is paid to constrained problems, overlapping problems and computationally expensive problems. The main innovations of the newly proposed algorithms focus on the decomposition of the large-scale problems and the cooperation between the obtained subproblems. Numerical experiments are conducted to put forward the benefits of the developed tools.
|Date of Award||29 Jun 2021|
|Sponsors||Université de Namur|
|Supervisor||Timoteo Carletti (Supervisor), ANNICK SARTENAER (Jury), Charlotte BEAUTHIER (Jury), Alexandre Mayer (Jury), Daniel Tuyttens (Jury) & Mohammed El-Abd (Jury)|