FORMATION OF REQUIREMENTS FOR THE SELF-ADAPTIVE GENETIC ALGORITHM

Abstract

Genetic algorithms (GAs) are proficient problem solvers, yet their effectiveness 
hinges on well-defined parameter settings that often vary by problem domain. This limita
tion has given rise to Self-Adaptive Genetic Algorithms (SAGAs) capable of autonomously 
fine-tuning of their parameters. Our review studies the spectrum of adaptation techniques 
enabling SAGAs to dynamically optimize their performance. We discern two categories of 
adaptation approaches: those inherently related to specific problems and more versatile 
strategies applicable across the domains. This dichotomy highlights the need for intelligent 
detection algorithms to distinguish when to employ problem-specific or generalized adapta
tions. The array of adaptation techniques presents a challenge in selecting the most suitable 
adaptations. This underscores the need for algorithms capable of making these choices, 
albeit with consideration for the computational and temporal demands of sophisticated ad
aptations. To unlock SAGAs' full potential, recognizing problem-specific adaptations is 
crucial, requiring detectors capable of discerning problem characteristics. Coexisting adap
tations may conflict, necessitating resolution within SAGAs. Temporal dynamics also play 
a role, with some adaptations suited for short iterations and others for extended conver
gence periods. SAGAs must navigate these nuances to optimize efficiency. In summary, 
our review navigates the realm of self-adaptive genetic algorithms, offering insights into 
their potential, challenges, and key considerations, ultimately contributing to the develop
ment of powerful and versatile optimization tools.