A Novel Digital Realization of AdEx Neuron Model

Abstract

Realization of biological neuron models is an essential research area in the field of neuromorphic. This brief presents a novel implementation of adaptive-exponential (AdEx) neuron model based on a complete-matching method called Power-2 Based AdEx Model (PBAM). This model can precisely reproduce different spiking behaviors, similar to the biological neurons with lower implementation costs compared with previous works. To validate the PBAM neuron, the proposed model is physically realized on FPGA as a proof of concept. Experimental results demonstrate high similarity with the original model, high computational performance and lower hardware cost. The proposed PBAM neuron implementation on FPGA requires considerably lower hardware resources compared with the original AdEx neuron. In comparison with similar works, this model has a higher performance and can operate at higher frequencies. In this modification, time domain spiking and dynamical behaviors of the original model are regenerated with very low computational errors using low-cost fixed-point calculations. This makes the proposed model an ideal candidate for large scale neuromorphic and biologically inspired neural network implementations targeting low-cost hardware platforms.