Perspectives on Superconducting Neural Networks

This is the authors' abstract. We don't add key points for this paper.

Neural networks are a computational method mimicking animal brains and are characterized by their massively parallel computation capabilities. Their basic components are neurons, which generate output when input exceeds a certain threshold, and synapses, which weight the connections between neurons. A neural network is a massively parallel circuit in which numerous neurons are connected via numerous synapses. Considering that a neuon in the human brain is connected with hundreds to tens of thousands of neurons via synapses, achieving something similar in an electrical circuit requires high signal transmission capacity and low power consumption. Superconducting devices are considered a promising solution. Superconducting wiring can transmit signals over long distances without attenuation. Furthermore, their superconductivity eliminates power consumption. This paper traces the history of superconducting neural network development and introduces the latest research trends for the future.