Adaptive Distributed Differential Evolution.

Area of Science:

• Computational intelligence
• Optimization algorithms
• Evolutionary computation

Background:

• Distributed Differential Evolution (DDE) is a powerful global optimization technique.
• DDE struggles with sensitive strategy selection and parameter settings, hindering performance.
• Existing DDE variants often lack adaptive mechanisms to address these challenges.

Purpose of the Study:

• To propose an Adaptive Distributed Differential Evolution (ADDE) algorithm.
• To enhance the robustness and performance of DDE by adaptively managing strategies and parameters.
• To address the sensitivity of DDE to strategy selection and parameter configuration.

Main Methods:

• Introduced a master-slave multipopulation distributed framework with three co-evolved populations: exploration, exploitation, and balance.
• Implemented adaptive mutation strategy selection based on evolutionary state estimation for each population.
• Developed adaptive updates for individual parameters (amplification factor F, crossover rate CR) and population size (N) using historical success and solution improvement.

Main Results:

• ADDE demonstrated significant superiority over state-of-the-art DDE and adaptive differential evolution variants.
• Performance was validated on 30 benchmark functions (CEC 2014) and 22 real-world problems (CEC 2011).
• The adaptive mechanisms effectively reduced sensitivity to strategy and parameter choices.

Conclusions:

• The proposed ADDE algorithm offers a robust and effective solution for complex global optimization problems.
• Adaptive strategy and parameter tuning are crucial for enhancing DDE performance.
• ADDE provides a promising advancement in distributed evolutionary optimization.