Revealing Unique Scaling Effects of Random Telegraph Noise and Electron Injection Stochasticity in Stochastic Resonance with Floating Gate-based Neurons

Random telegraph noise (RTN) during reading and electron injection stochasticity (EIS) during integration are integrated and applied to a summing network with floating gate (FG)-based neurons for the first time. The stochastic resonance (SR) phenomena are simulated to segment the influence of RTN and EIS. The comparative analysis of RTN and EIS reveals the unique advantages of EIS on SR originating from its high probability and input bias dependent stochastic characteristics. Thanks to these characteristics, EIS demonstrates 70x better signal detection error rate over RTN in simulation. It is found that the EIS is suitable source of the stochasticity to enhance the SR effects in the summing network, paving the way for the realization of a FG-based neuron as a brain-inspired computational primitive.