Stochastic Resonance Effects of Floating Gate Technology-based Leaky Integrate-and-Fire (FG LIF) Neurons in Summing Network

The stochastic resonance effects are successfully simulated for the floating gate technology-based leaky integrate-and-fire (FG LIF) neurons in a summing network. Thanks to the stochastic electron injection effects of FG LIF neurons, the network can detect very short and weak input pulses. With the aggressively scaled FG LIF neurons from 100nm to 10nm, the representation error is reduced from 28% to 0.3% owing to the increased stochasticity of electron injections. In addition, the impacts of the leaky characteristics of FG LIF neurons on the representation error rate are discovered and discussed.