1T FDSOI Based LIF Neuron Without Reset Circuitry: A proposal and Investigation

Abstract

In this article, an FDSOI-based neuron is demonstrated to mimic the functionalities of LIF neuron. As the proposed device-based neuron utilizes the single transistor latch (STL) mechanism and it takes the current input and voltage output in the shape of a spike, it does not require any extra circuit for reset. Furthermore, the proposed neuron shows an impressive energy consumption per spike i.e., less than 0.11 pJ/spike for all input values under 500 nA. Interestingly, it shows a spiking frequency in MHz range, which is ~5 orders greater than the biological neuron. Hence, due to its high frequency, energy and area efficiency, equivalent to that of a biological neuron and compatibility with CMOS process, the proposed FDSOI-based LIF neuron is more appropriate for large-scale hardware implementation of a spiking neural network (SNN).