Superconducting Neural Networks: from an Idea to Fundamentals and, Further, to Application

• Published in: Nanobiotechnology Reports (Online) Vol. 16; no. 6; pp. 811 - 820
• Main Authors: Schegolev, A. E., Klenov, N. V., Soloviev, I. I., Gudkov, A. L., Tereshonok, M. V.
• Format: Journal Article
• Language: English
• Published: Moscow Pleiades Publishing 01.11.2021; Springer Nature B.V
• Subjects: Algorithms; Artificial neural networks; Chemistry and Materials Science; Complex systems; Global optimization; Industrial and Production Engineering; Machines; Manufacturing; Materials Science; Microprocessors; Nanoelectronics and Neuromorphic Computer Systems; Nanotechnology; Neural networks; Object recognition; Processes; Quantum computing; Speech recognition; Superconductivity
• ISSN: 2635-1676; 1995-0780; 2635-1684; 1995-0799
• DOI: 10.1134/S2635167621060227

The popularity and diversity of artificial neural networks for various applications are ever increasing. The development of neural networks in the form of software models and hardware systems emphasizes their relevance and range of applicability, from a ten-minute Python code, an AlphaZero neural network, and intelligent image and speech recognition algorithms to IBM and Qualcomm neuromorphic chips and D-Wave quantum computing systems. The superconductor implementation of neural networks, along with the obvious advantages of superconductor technology in terms of energy efficiency and operating speed, makes it possible to combine a neural network and a superconducting quantum processor in one computing unit. In this case, the quantum core of a complex system can be used to learn a neural network by a global optimization method. It is noteworthy that the world’s leading IT companies clearly demonstrate the market’s focus on superconducting elements. The relevance of this direction is analyzed against a historical retrospective.