Can genetic algorithms evolve without mutation? This paper explores the self-adaptive capabilities of real-parameter genetic algorithms (GAs) using a simulated binary crossover (SBX) operator, surprisingly achieving this without any mutation operator. It highlights the essential feature of self-adaptation found in natural evolution and seeks to implement it within function optimization. The study draws a connection between the workings of self-adaptive evolution strategies (ESs) and real-parameter GAs with SBX. It then demonstrates the self-adaptive behavior of real-parameter GAs on a series of test problems commonly used in the ES field. The problems provide a benchmark for comparing the performance of different algorithms. Ultimately, the findings reveal a remarkable similarity in the working principles of real-parameter GAs and self-adaptive ESs. This similarity suggests the need for further investigation into self-adaptive GAs. This research contributes to a deeper understanding of evolutionary search algorithms and opens new avenues for exploration in the field of genetic algorithms.
Appearing in Evolutionary Computation, this research directly addresses the journal’s core theme of computational methods inspired by natural evolution. The paper’s focus on self-adaptive genetic algorithms contributes to the ongoing development of optimization techniques within the evolutionary computation paradigm. It is relevant to other works published in the journal that explore genetic algorithms and evolutionary strategies.