A comparison study of self-adaptation in evolution strategies and real-coded genetic algorithms

Area of Science:

• Evolutionary Computation
• Optimization Algorithms

Background:

• Continuous search spaces present challenges for optimization algorithms.
• Self-adaptive mechanisms enhance the performance of evolutionary algorithms.

Purpose of the Study:

• To compare the self-adaptive mechanisms of evolution strategies (ES) and real-coded genetic algorithms (RCGA).
• To analyze the adaptive characteristics of these algorithms in continuous optimization.

Main Methods:

• Numerical experiments were conducted to evaluate ES and RCGA.
• The study examined self-adaptive parameters in ES, such as mutation standard deviations.
• RCGA's adaptive offspring generation using recombination was analyzed.

Main Results:

• ES utilizes self-adaptive mutation parameters that evolve alongside decision variables.
• RCGA employs recombination to generate offspring adaptively without explicit adaptive parameters.
• Both algorithms exhibit distinct self-adaptive characteristics like distribution translation, enlargement, focusing, and directing.

Conclusions:

• Evolution strategies and real-coded genetic algorithms offer different approaches to self-adaptation in continuous optimization.
• Understanding these mechanisms is crucial for selecting appropriate algorithms for specific optimization tasks.