[14a-B408-1~10] 11.5 Junction and circuit fabrication process, digital applications

[14a-B408-8] Design of neuron with sigmoidal activation functions for superconducting neural network

〇Daiki Yamaguchi¹, Yuki Yamanashi¹, Nobuyuki Yoshikawa¹ (1.Dept. of Electrical and Computer Eng., Yokohama Natl. Univ.)

Keywords:artificial neural network

In this study, we designed a neuron with hyperbolic tangent activation function. The designed neuron can change the threshold by applying a bias. It is possible to learn ANN by changing the threshold of neurons. A simple ANN was constructed using the neurons designed in this study, and the circuit was investigated. The output of the configured ANN can be controlled by changing the bias of each neuron.