Currently, studies in an area of Artificial Intelligence are among the most relevant and promising ones in the scientific environment. At the same time, issues and problems on this topic are examined in various scientific areas: philosophy, mathematics, technical disciplines, and many others. A growing interest in this subject was caused by the scientific breakthrough of the 20th century in an area of neurophysiology and neuroanatomy, and, on the basis of it, mathematical and hardware-technical models of artificial neural networks were developed. This tool largely allowed simulating various aspects of human thinking. Therefore, modern computers acquired many human-like functions that led to a significant increase of scientific and technological progress and, in fact, launched a global revolution of universal digitalization. In this regard, the purpose of the article is an attempt to systematize knowledge on existing achievements in the area of artificial neural network modeling and its shortcomings in the solution of various types of applied problems, which, in turn, will allow specialists of different scientific areas to successfully use this tool in their work and to navigate in its limitations. The scientific novelty of the review follows the need for regular monitoring of this theme due to frequent breakthroughs and successes in this area. Currently available architectures and topologies of artificial neural networks, which received wide and efficient applications in modern human activities (channels of direct distribution, convolutional and recurrent neural networks), are presented in the paper. Primary shortcomings of models, identified after a comparison with an actual work of a human brain, characterizing a thought process and intelligence, which is also a significant constraint in the creation of a full-fledged human-like computer intellectual system, are also analyzed. General scientific methods, such as generalization, analysis, systematization, etc., were used in this research
Keywords
classification, artificial neural networks, artificial intelligence, thinking, neurophysiology, pattern recognition