Neuron Circuit Using Coupled SQUIDs Gate with Flat Output Characteristics for Superconducting Neural Network

Abstract

We propose an improved design of a neuron circuit, using coupled SQUIDs gates, for a superconducting neural network. An activation function with step-like input vs. output characteristics is desirable for a neuron circuit to solve a combinatorial optimization problem. The proposed neuron circuit is composed of two coupled SQUIDs gates with a cascade connection, in order to obtain such characteristics. The designed neuron circuit is fabricated by a 2.5kA/cm² Nb/AlOx/Nb process. The operation of a fabricated neuron circuit is experimentally demonstrated. Network performance of a neural network using proposed neuron circuits is also estimated by numerical dynamic simulations.