Major achievements in designing and analyzing recurrent spiking neural networks intended for modeling functional brain networks are reviewed. Key terms and definitions employed in machine learning are introduced. The main approaches to the development and exploration of spiking and rate neural networks trained to perform specific cognitive functions are presented. State-of-the-art neuromorphic hardware systems simulating information processing by the brain are described. Concepts of nonlinear dynamics are discussed, which enable identification of the mechanisms used by neural networks to perform target tasks.

Keywords: artificial neural networks, nonlinear dynamics, machine learning, spiking neurons, modeling of cognitive functions PACS: 07.05.Mh, 84.35.+i, 87.19.L− (all) DOI: 10.3367/UFNe.2021.08.039042 URL: https://ufn.ru/en/articles/2022/10/b/ 001112536300002 2-s2.0-85182910380 2022PhyU...65.1020M Citation: Maslennikov O V, Pugavko M M, Shchapin D S, Nekorkin V I "Nonlinear dynamics and machine learning of recurrent spiking neural networks" Phys. Usp. 65 1020–1038 (2022) BibTexBibNote ® (generic)BibNote ® (RIS)MedlineRefWorks Received: 1st, June 2021, revised: 13th, August 2021, accepted: 13th, August 2021 Оригинал: Масленников О В, Пугавко М М, Щапин Д С, Некоркин В И «Нелинейная динамика и машинное обучение рекуррентных спайковых нейронных сетей» УФН 192 1089–1109 (2022); DOI: 10.3367/UFNr.2021.08.039042