부산대학교 도서관

The main challenge for the hardware implementation of spiking neural networks is the design of a reliable neuron. Soma, which is the nucleus of the neuron, is a key part of such a design. More precisely, the soma design must accurately capture the excitatory/inhibitory interactions, intrinsic charge dynamics, refractory period, spike encoding, neuronal action potential, and output spike firing processes in order to mimic the corresponding biological processes. This work presents the design of an artificial soma cell with excitatory and inhibitory inputs, called $\upalpha$-Soma. The key idea is to design and employ a new interconnect cell, named $\upalpha$-cell, and integrate this cell within the proposed soma cell design. The design implementation utilizes the Rapid Single Flux Quantum (RSFQ) logic circuit technology. We demonstrate the correct functionality, high performance, and energy efficiency of the proposed $\upalpha$-Soma cell with detailed circuit simulations under different spiking conditions.