Numerical Investigation of the Swarm Intelligence Algorithm Obtained Using ChatGPT for Univariate Global Optimization

This paper explores the possibility of designing an efficient global optimization algorithm using an artificial intelligence chatbot, ChatGPT. The main idea is to use the swarm intelligence metaheuristic method, which operates on a set of particles in a swarm, to solve non-local univariate search problems. Testing was carried out on a set of one-dimensional non-convex functions with varying numbers of local optima. The results demonstrate that this algorithm can be used to solve multi-extremal optimization problems. The paper presents the results of a numerical study on the properties of this algorithm in comparison with the Piyavsky method, the Strongin method, and a combination of the Strongin with the Parabolas methods. The performed computational experiments testify to the suitability of the obtained swarm intelligence algorithm for solving univariate non-local search problems. This approach can also be useful for solving auxiliary optimization problems and for developing multimethod computational technologies.