Digital control of a superconducting adiabatic sigma neuron

Superconducting adiabatic neurons based on quantum parametrons make it possible to “calculate” sigmoid activation functions and their derivatives in an analog manner in one cycle with record energy efficiency. However, for the application of such systems, it is necessary to determine operation regimes that allow integration with Josephson digital circuits, where information is presented in the form of sequences of picosecond unipolar voltage pulses. The conditions allowing the pulse control of an adiabatic sigma neuron have been theoretically determined in this work. The obtained results allowed us to develop the design of the basic unit for the hardware implementation of analog–digital Josephson neural networks.