GENETIC ALGORITHMS AS A METHOD OF GLOBAL OPTIMIZATION: POSSIBILITIES AND LIMITATIONS

ABSTRACT

This article discusses the basic principles and structure of genetic algorithms, an optimization method based on the principles of natural selection and genetic evolution. The key stages of genetic algorithms, including initialization, evaluation, selection, crossover, and mutation, are described, and a flow chart of their operation is provided. The application areas of GA, such as industrial optimization, machine learning, bioinformatics, and financial analysis, are analyzed, where this method has proven itself as an effective tool for solving complex problems with large search spaces. The advantages of genetic algorithms, such as global search, flexibility, and the ability to run in parallel, are noted, as well as their limitations related to computational costs and parameter dependence. The conclusion emphasizes that genetic algorithms, if properly configured, can be indispensable for problems requiring adaptability and high accuracy.

АННОТАЦИЯ

В данной статье рассматриваются основные принципы и структура генетических алгоритмов — метода оптимизации, основанного на принципах естественного отбора и генетической эволюции. Описаны ключевые этапы работы генетических алгоритмов, включая инициализацию, оценку, отбор, кроссинговер и мутацию, а также приводится схема их работы. Проанализированы области применения ГА, такие как промышленная оптимизация, машинное обучение, биоинформатика и финансовый анализ, где этот метод зарекомендовал себя как эффективный инструмент для решения сложных задач с большими пространствами поиска. Отмечены преимущества генетических алгоритмов, такие как глобальный поиск, гибкость и возможность параллельного выполнения, а также их ограничения, связанные с вычислительными затратами и зависимостью от параметров. Заключение подчеркивает, что генетические алгоритмы, при правильной настройке, могут быть незаменимыми для задач, требующих адаптивности и высокой точности.

Keywords: genetic algorithms, optimization, evolutionary computation, fitness function, crossing over, mutation, selection, adaptive algorithms, global search, multidimensional search spaces.

Ключевые слова: генетические алгоритмы, оптимизация, эволюционные вычисления, фитнес-функция, кроссинговер, мутация, отбор, адаптивные алгоритмы, глобальный поиск, многомерные пространства поиска.

Genetic algorithms (GA) are one of the optimization and solution search methods based on the principles of biological evolution, such as natural selection, heredity, and mutation. These algorithms belong to the class of evolutionary computations and represent an approach in which the solution to a problem is modeled using a population of possible solutions. The approach imitates the process of natural evolution, gradually improving the population of solutions and approaching the optimal solution to the problem [1]. The principles of GAs were proposed by American scientist John Holland in the 1960s. In his book Adaptation in Natural and Artificial Systems (1975), Holland described the theoretical basis for algorithms that can “learn” and “evolve” using mechanisms such as heredity and mutation. These processes allow an algorithm to improve its solution over time by applying random but controlled changes to existing solutions to create new, potentially more efficient solutions.

Genetic algorithms (GAs) are versatile tools used across various fields to solve problems with large, complex search spaces. In industry, GAs optimize processes, enhance productivity, and reduce costs by improving operation sequencing, production parameters, and logistics. In machine learning and AI, they efficiently tune hyperparameters and optimize neural network architectures, while in robotics, they enhance movement, trajectory, and behavior strategies for dynamic adaptation.

Beyond these areas, GAs play a crucial role in bioinformatics, gaming, and energy management. They aid genome research, protein structure prediction, and sustainable energy strategies by optimizing resource usage. In gaming, they create adaptive content and improve NPC behaviors, showcasing their effectiveness in addressing diverse and complex challenges.

GA is an iterative process that uses the principles of natural selection and genetics to find optimal solutions (Fig.1). Its structure includes several key stages: initialization, evaluation, selection, crossing over, mutation and replacement. Each of these stages is necessary to form a new generation of solutions and get closer to the optimal solution to the problem. Let's consider each stage in more detail and present the general algorithm in the form of a flowchart.

Figure 1. Block diagram of the genetic algorithm

The main stages of the genetic algorithm structure are: initialization, populations, fitness function (fitness assessment), selection of the best solutions, crossing over, mutation, replacement (population update). [5].

The genetic algorithm (GA) follows several main stages: population initialization, fitness function calculation, selection, crossover, mutation, population replacement, and applying a stopping criterion. The process starts by creating an initial population of potential solutions, evaluating their fitness, and selecting the fittest individuals for reproduction. Through crossover and mutation, new solutions are generated, allowing the population to evolve iteratively toward optimal or near-optimal solutions. This systematic approach mimics natural selection and adaptation, making GAs powerful for solving complex optimization problems.

GAs have notable advantages, including global search capability, flexibility, and suitability for large search spaces with complex constraints. They do not rely on derivatives or gradients and are parallelizable, making them efficient for diverse applications. However, they also have limitations, such as high computational costs, sensitivity to parameter settings, and the risk of premature convergence. Additionally, the algorithm's success depends heavily on the design of the fitness function, and there is no absolute guarantee of finding the optimal solution. These factors should be carefully considered when selecting GAs for specific tasks.

GAs are a powerful and flexible optimization method based on the principles of biological evolution, such as natural selection, crossover, and mutation. These algorithms allow finding optimal or near-optimal solutions in complex and multidimensional search spaces, where traditional methods often fail. Due to their versatility, GAs are used in a variety of fields, from industrial optimization and machine learning to financial analysis and bioinformatics.

Among the advantages of genetic algorithms are their ability to find global optimal solutions, versatility and flexibility, as well as the absence of the need to calculate derivatives. However, GAs also have their limitations: they require significant computational resources, are sensitive to parameters, and depend on the fitness function.

References:

1. Kumar, M., Husain, D., Upreti, N., & Gupta, D. Genetic Algorithm: Review and Application. Electrical Engineering eJournal. https://doi.org/10.2139/ssrn.3529843
2. Radenbaugh, A. Applications of genetic algorithms in bioinformatics. https://doi.org/10.31979/etd.m44r-26hr
3. Merino, D., Reyes, E., & Steidley, C. Genetic Algorithms: Theory And Application. https://doi.org/10.18260/1-2--7144.
4. Алиев Р.А., Алиева Р.Р. Генетические алгоритмы и их применение в задачах оптимизации. Москва: МГУ. 2021. 320 стр.
5. Эйхвальд И.Я., Шитиков В.К. Генетические алгоритмы и оптимизация: основы теории и приложения. СПб.: Питер. 2013. 288 стр.
6. Ф. Хошимов, И.У. Рахмонов. Методы расчета прогнозных значений норм удельного электропотребления на предприятиях с меняющейся величиной потребляемой мощности // Universum: технические науки : электрон. научн. журн. 2021. 10(91). URL: https://7universum.com/ru/tech/ archive/item/12384