Structure of associative heterarchical memory

Objectives. Since the 20th century, artificial intelligence methods can be divided into two paradigms: top-down and bottom-up. While the methods of the ascending paradigm are difficult to interpret as natural language outputs, those applied according to the descending paradigm make it difficult to actualize information. Thus, natural language processing (NLP) by artificial intelligence remains a pressing problem of our time. The main task of NLP is to create applications that can process and understand natural languages. According to the presented approach to the construction of artificial intelligence agents (AI-agents), processing of natural language should be conducted at two levels: at the bottom, methods of the ascending paradigm are employed, while symbolic methods associated with the descending paradigm are used at the top. To solve these problems, the authors of the present paper propose a new mathematical formalism: associative heterarchical memory (AH-memory), whose structure and functionality are based both on bionic principles and on the achievements of top-down and bottom-up artificial intelligence paradigms.

Methods. Natural language recognition algorithms were used in conjunction with various artificial intelligence methods.

Results. The problem of character binding as applied to AH-memory was explored by the research group in earlier research. Here, abstract symbol binding was performed using multi-serial integration, eventually converting the primary symbols produced by the program into integrated abstract symbols. The present paper provides a comprehensive description of AH-memory in the form of formulas, along with their explanations and corresponding schemes.

Conclusions. The most universal structure of AH-memory is presented. When working with AH-memory, a developer should select from a variety of possible module sets those AH-memory components that support the most successful and efficient functioning of the AI-agent.

1. Dushkin R.V. Iskusstvennyi intellekt (Artificial intelligence). Moscow: DMK-Press; 2019. 280 p. (in Russ.). ISBN 978-5-97060-787-9

2. Dushkin R.V. Review of artificial intelligence approaches and methods. Radioelektronnye tekhnologii. 2018;3: 85-89 (in Russ.).

3. Nikolenko S., Arkhangel'skaya E., Kadurin A. Glubokoe obuchenie. Pogruzhenie v mir neironnykh setei (Deep learning. Immersion in the world of neural networks). St. Petersburg: Piter; 2018. 480 p. (in Russ.). ISBN 9785-496-02536-2

4. Sarker M.K., Zhou L., Eberhart A., Hitzler P. Neuro-symbolic artificial intelligence: Current trends. AI Communications. 2021. 13 p. https://doi.org/10.48550/arXiv.2105.05330

5. Raina V., Krishnamurthy S. Natural language processing. In: Building an Effective Data Science Practice. Apress, Berkeley, CA; 2022. https://doi.org/10.1007/978-1-4842-7419-4_6

6. Zadeh L.A. From computing with numbers to computing with words - From manipulation of measurements to manipulation of perceptions. Int. J. Appl. Math. Comput. Sci. 2002;12(3):307-324. Available from URL: https://www.zbc.uz.zgora.pl/Content/2928/1zade.pdf

7. Dushkin R.V. Development of adaptive learning methods using intelligent agents. Iskusstvennyi intellekt i prinyatie reshenii. 2019;1:87-96 (in Russ.). https://doi.org/10.14357/20718594190108

8. Zhu L., Gao W. Hypergraph ontology sparse vector representation and its application to ontology learning. In: Tan Y., Shi Y., Zomaya A., Yan H., Cai J. (Eds.). Data Mining and Big Data. International Conference on Data Mining and Big Data (DMBD 2021). Part of the Communications in Computer and Information Science. Book series. Singapore: Springer; 2021. V. 1454. P. 16-27. https://doi.org/10.1007/978-981-16-7502-7_2

9. Kormen T., Leizerson Ch., Rivest R., Shtain K. Algoritmy. Postroenie i analiz. Moscow: Vil'yams; 2011. 1296 p. (in Russ.). ISBN 0-07-013151-1 [Cormen T.H., Leiserson C.E., Rivest R.L., Stein C. Introduction to Algorithms. London: The MIT Press; 2009. 1312 p.]

10. Horiguchi H., Winawer J., Dougherty R.F., Wandell B.A. Human trichromacy revisited. Proceedings of the National Academy of Sciences (PNAS). 2012;110(3):E260-E269. https://doi.org/10.1073/pnas.1214240110

11. Harnad S. The symbol grounding problem. Physica D: Nonlinear Phenomena. 1990;42(1-3):335-346. https://doi.org/10.1016/0167-2789(90)90087-6

12. Jentzsch R., Feustel D., Topf B. Unique identifier, method for providing the unique identifier and use of the unique identifier: US Pat. US8578162B2. Publ. 05.11.2013.

13. Kaminski B., Pralat P., Theberge F. Hypergraphs. In: Mining Complex Networks. NY: Chapman and Hall/CRC; 2021. https://doi.org/10.1201/9781003218869-7

14. Popov E.V. Obshchenie s EVM na estestvennom yazyke (Communication with a computer in natural language). Moscow: Nauka; 1982. 360 p. (in Russ.).

15. Malchukov A., De Swart P. Differential case marking and actancy variation. In: The Oxford Handbook of Case. Oxford: Oxford University Press; 2012.