An Efficient Multiple Variants Coordination Framework for Differential Evolution.

Area of Science:

• Computational intelligence
• Optimization algorithms
• Evolutionary computation

Background:

• Differential evolution (DE) is a powerful evolutionary algorithm, but its variants often excel only on specific function types.
• A single DE variant may not be optimal for the entire optimization process, limiting overall performance.

Purpose of the Study:

• To propose a Multiple Variants Coordination (MVC) framework to improve the performance of differential evolution algorithms.
• To address the limitations of single DE variants by coordinating multiple optimizers.

Main Methods:

• The proposed MVC framework divides the evolution process into segments, each with learning generations (LGs) and executing generations (EGs).
• In LGs, candidate DE optimizers are evaluated independently to select the best performer for the subsequent EG.
• The Multiple Variants Adaptive Solutions Preserving mechanism (MV-APM) adaptively preserves promising solutions from various optimizers.

Main Results:

• Numerical experiments on the CEC2014 benchmark suite demonstrate that the MVC framework significantly enhances baseline DE algorithm performance.
• The resulting algorithm using MVC outperforms state-of-the-art and up-to-date DE algorithms.
• The MVC framework shows general applicability for coordinating multiple improved DE variants.

Conclusions:

• The MVC framework effectively coordinates multiple DE variants, leading to superior optimization performance.
• MVC offers a generalizable approach to enhance the capabilities of existing and future DE algorithms.
• This framework provides a robust solution for complex optimization problems where single optimizers may falter.