Scaling up Multi-island Competitive Cooperative Coevolution for Real Parameter Global Optimisation

Autor:	Rohitash Chandra, Kavitesh Kumar Bali
Rok vydání:	2015
Předmět:	Constraint (information theory) Variable (computer science) Cooperative coevolution Decision variables Theoretical computer science Context vector Computer science Decomposition (computer science) Scaling Task (project management)
Zdroj:	AI 2015: Advances in Artificial Intelligence ISBN: 9783319263496 Australasian Conference on Artificial Intelligence
DOI:	10.1007/978-3-319-26350-2_4
Popis:	A major challenge in using cooperative coevolution (CC) for global optimisation is the decomposition of a given problem into subcomponents. Variable interaction is a major constraint that determines the decomposition strategy of a problem. Hence, finding an optimal decomposition strategy becomes a burdensome task as inter-dependencies between decision variables are unknown for these problems. In recent related work, a multi-island competitive cooperative coevolution (MICCC) algorithm was introduced which featured competition and collaboration of several different decomposition strategies. MICCC used five different uniform problem decomposition strategies that were implemented as independent islands. This paper presents an analysis of the MICCC algorithm and also extends it to more than five islands. We incorporate arbitrary (non-uniform) problem decomposition strategies as additional islands in MICCC and monitor how each different problem decomposition strategy contributes towards the global fitness over different stages of optimisation.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_________::a275507d063f4d6450f85c4aca508ee4 https://doi.org/10.1007/978-3-319-26350-2_4 Zobrazit plný text záznamu