Improved Constructive Cooperative Coevolutionary Differential Evolution for Large-Scale Optimisation
Paper in proceedings, 2015

The Differential Evolution (DE) algorithm is widely used for real-world global optimisation problems in many different domains. To improve DE's performance on large-scale optimisation problems, it has been combined with the Cooperative Coevolution (CCDE) algorithm. CCDE adopts a divide-and-conquer strategy to optimise smaller subcomponents separately instead of tackling the large-scale problem at once. DE then evolves a separate subpopulation for each subcomponent but there is cooperation between the subpopulations to co-adapt the individuals of the subpopulations with each other. The Constructive Cooperative Coevolution ((CDE)-D-3) algorithm, previously proposed by the authors, is an extended version of CCDE that has a better performance on large-scale problems, interestingly also on non-separable problems. This paper proposes a new version, called the Improved Constructive Cooperative Coevolutionary Differential Evolution ((CDE)-D-3i), which removes several limitations with the previous version. A novel element of (CDE)-D-3i is the advanced initialisation of the subpopulations. (CDE)-D-3i initially optimises the subpopulations in a partially co-adaptive fashion. During the initial optimisation of a subpopulation, only a subset of the other subcomponents is considered for the co-adaptation. This subset increases stepwise until all subcomponents are considered. The experimental evaluation of (CDE)-D-3i on 36 high-dimensional benchmark functions (up to 1000 dimensions) shows an improved solution quality on large-scale global optimisation problems compared to CCDE and DE. The greediness of the co-adaptation with (CDE)-D-3i is also investigated in this paper.

algorithm

spaces

global optimization

Computer Science

Author

E. Glorieux

University West

B. Svensson

University West

F. Danielsson

University West

Bengt Lennartson

Chalmers, Signals and Systems, Systems and control, Automation

Proceedings - 2015 IEEE Symposium Series on Computational Intelligence, SSCI 2015

1703-1710

Subject Categories

Computer Vision and Robotics (Autonomous Systems)

DOI

10.1109/ssci.2015.239

More information

Latest update

1/27/2021