Visual Analysis of Leaky Integrate-and-Fire Spiking Neuron Models and Circuits

Emulating biologically plausible online learning in spiking neural networks (SNNs) will enable the next generation of energy-efficient neuromorphic architectures. While software leads the way in terms of exploring various Machine Learning (ML) algorithms and applications, bridging the gap between hardware (devices and circuits) and software is crucial to accurately predict network properties, especially at large scale. This work compares behavior of a spiking neuron circuit simulated with Cadence Spectre to a Python model implemented with a custom spiking neuron model. The results demonstrate that the two exhibit the same spiking characteristics over a range of parameter values, confirming that the more versatile Python model indeed has a hardware equivalent.