Hopfield neural network as an open dynamic system
DOI:
https://doi.org/10.34121/1028-9763-2022-4-3-14Keywords:
open dynamic system, Hopfield neural network, convergence, відкрита динамічна система, нейромережа Хопфілда, конвергенціяAbstract
Abstract. The paper provides an analysis of a Hopfield neural network model from the standpoint of the concept of open dynamic systems (ODS) offered earlier. According to it, the evolution of the universe can be represented as a sequence of stages of the ODS development, during which more and more complex ODSs were formed. Living cells were later formed from them and their further development led to the appearance of humans and their intelligence, the formation of civilization, and the emergence of artificial intelligence. An overview of the provisions of the ODS concept is presented and the stage of the emergence of life as a collective development of ODS in the form of populations of the offspring of the mother cell, that led to the emergence of complex organisms, the nervous system, human, and intelligence as an intangible form of the ODS development, is analyzed in detail. It has been suggested that transforming an inanimate macromolecule of ODS into a living cell is connected with the appearance in its composition of a multistable component like a Hopfield network, whose prototype – spin glass – is a quantum object that existed even before the emergence of life on Earth. To verify it, there has been carried out a detailed analysis of the Hopfield network model as a linear manifold stretched over a set of memorized data vectors. New evidence of the network convergence process coinciding is presented, and the source of energy that a living cell requires for reproduction is revealed. In the paper, a structure of a living ODS that is similar to the Hopfield network and supplemented with a reproduction module similar to a DNA molecule is offered. It fully corresponds to the structure of the first bacteria on Earth detected by archaeologists. The results of the study confirm the efficiency of the methods of the ODS concept when studying the dynamics of the development of systems of different natures, including the latest artificial intelligence systems whose behavior may not be controlled.References
1. Rosenblatt F. The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain. Psychol. Rev. 1958. N 65. Р. 386–408.
2. Розенблатт Ф. Принципы нейродинамики (перцептроны и теория механизмов мозга). М.: Мир, 1965. 473 с.
3. Hebb D.O. The organization of behavior. J. Wiley, 1949.
4. Минский М., Пейперт С. Персептроны. М.: Мир, 1971. 252 с.
5. Hopfield J.J. Neural networks and physical systems with emergent collective computational abilities. Proc. Nat. Acad. Sci. U.S.A. 1982. N 79 (8). P. 2554–2558.
6. Kirkpatrick S., Sherrington D. Infinite-ranged models of spin-glasses. Physical Review. 1978. Vol. 17. Series B. P. 4384–4403.
7. Хайкин С. Нейронные сети. Полный курс. М.: Вильямс, 2006. 1103 с.; Reed R.D., Marks II R.J. Neural Smithing, MIT Press, 1995.
8. Rumelhardt D.E., Hinton G.E., Williams R.I. Learning Internal Representations by Error Propagation. Parallel Distributed Processing: Explorations in the Microstructure of Cognition. Cambridge, Mass.: MIT Press. 1986. Vol. 1. P. 318–362.
9. Krotov D., Hopfield J.J. Dense associative memory for pattern recognition / D.D. Lee, M. Sugiyama, U.V. Luxburg, I. Guyon, R. Garnett (eds.). Advances in Neural Information Processing Systems. 2016. N 29. P. 1172–1180.
10. Різник О.М. Концепція відкритої динамічної системти. Математичні машини і системи. 2020. № 3. С. 3–22.
11. Месарович М., Такахара Я. Общая теория систем. Математические основы. М.: Мир, 1978. 311 с.
12. Ейнштейн А. Основы общей теории относительности. Альберт Эйнштейн и теория гравитации. М.: Мир, 1979. С. 146–198.
13. Румер Ю.Б., Рывкин М.С. Теория относительности. М: ГУПИ МП. РСФСР, 1960. С. 212.
14. Месарович М. Основания общей теории систем. Общая теория систем / пер. с англ. М.: Мир, 1966. С. 15–48.
15. Уиллер Дж. Гравитация, нейтрино и вселенная. М., 1962.
16. Никонов О.А. Философские аспекты геометродинамики. Вестник МГТУ. 2011. Т. 14, № 2. С. 272–280.
17. Эткинс П. Кванты. Справочник концепций / пер. с англ. В.Л. Ядровского. М.: Мир., 1977. 496 с.
18. Майр Э. Эволюция. М.: Мир, 1981. 264 с.
19. Дикерсон Р.Е. Химическая эволюция и происхождение жизни. Эволюция. М.: Мир, 1981. C. 67–108.
20. Валентайн Дж.У. Эволюция многоклеточных организмов и животных. Эволюция. М.: Мир, 1981. С. 149–172.
21. Personnaz I., Guyon I., Dreyfus G. Collective computational properties of neural networks: New learning mechanisms. Phys. Rev. A. 1986. Vol. 34, N 5. P. 4217–4228.
22. Алберт А. Регрессия, псевдоинверсия и рекуррентное оценивание / пер. с англ. М.: Наука, 1977. 237 с.
23. Gorodnichy D.O., Reznik A.M. Increasing Attraction of Pseudo-Inverse Neural Networks. Neural Processing Letters 5. 1997. Р. 121–125.
24. Кириченко Н., Резник А., Щетинюк С. Псевдообращение матриц в проблеме проектирования ассоциативной памяти. Кибернетика и системный анализ. 2001. № 3. С. 18–27.
25. Резник А.М. Хопфилдовские ансамбли в латеральных нейроструктурах коры мозга. Математичні машини і системи. 2006. № 1. С. 3–12.
26. Reznik A.M., Sitchov A.S., Dekhtyarenko O.K., Nowicki D.W. Associative Memories with "Killed" Neurons: the Methods of Recovery. Proc. of the International Joint Conference on Neural Networks, (July 20–24, 2003, Portland, Oregon).
27. Кун Т. Структура научных революций. М.: Прогресс, 1975. 246 с.
183
79Published
Issue
Section
License
Copyright (c) 2022 Mathematical Machines and Systems
This work is licensed under a Creative Commons Attribution 4.0 International License.
