Cooperative co-evolution is recognized as an effective approach for solving large-scale optimization problems. It breaks down the problem dimensionality by splitting a large-scale problem into ones focusing on a smaller number of variables. This approach is successful when the studied problem is decomposable. However, many practical optimization problems can not be split into disjoint components. Most of them can be seen as interconnected components that share some variables with other ones. Such problems composed of parts that overlap each other are called overlapping problems. This paper proposes a modified cooperative co-evolutionary framework allowing to deal with non-disjoint subproblems in order to decompose and optimize overlapping problems efficiently. The proposed algorithm performs a new decomposition based on differential grouping to detect overlapping variables. A new cooperation strategy is also introduced to manage variables shared among several components. The performance of the new overlapped framework is assessed on large-scale overlapping benchmark problems derived from the CEC’2013 benchmark suite and compared with a state-of-the-art non-overlapped framework designed to tackle overlapping problems.
|Title of host publication||Optimization and Learning - 4th International Conference, OLA 2021, Proceedings|
|Subtitle of host publication||4th International Conference, OLA 2021, Catania, Italy, June 21-23, 2021, Proceedings|
|Editors||Bernabé Dorronsoro, Patricia Ruiz, Lionel Amodeo, Mario Pavone|
|Number of pages||13|
|Publication status||Published - 17 Aug 2021|
|Name||Communications in Computer and Information Science|
- Cooperative co-evolution
- Evolutionary algorithms
- Large-scale global optimization
- Overlapping problem
FingerprintDive into the research topics of 'Investigating Overlapped Strategies to Solve Overlapping Problems in a Cooperative Co-evolutionary Framework'. Together they form a unique fingerprint.
Challenging High Dimensionality in Evolutionary Optimization using Cooperative Co-evolutionary AlgorithmsAuthor: Blanchard, J., 29 Jun 2021
Student thesis: Doc types › Doctor of SciencesFile