Research Article

An Area-Efficient Integrate-and-Fire Neuron Circuit with Enhanced Robustness against Synapse Variability in Hardware Neural Network

Table 1

Comparison among the performances of reported neuron circuits and the proposed one in this work.

ReferencesSynaptic inputNeuron modelCircuit typeTechnology node (µm)Firing frequencyPower consumption (μW)Special features

[18]CurrentConductance-basedAnalog∼300 Hz60Neuro-physiological principles
[23]CurrentIntegrate-and-fireAnalog1.5∼1.2 kHz1.5Spike frequency adaptation
[26]CurrentHindmarsh–RoseMixed mode0.25∼2 kHz163.4Time-scaling techniques
[27]CurrentIzhikevichAnalog0.35∼1 MHz40Firing pattern and shape
[28]CurrentIntegrate-and-fireMixed mode0.8∼100 Hz120Spike-based learning
[29]VoltageIntegrate-and-fireMixed mode∼1 kHzPositive feedback neuron
This workCurrentLeaky integrate-and-fireAnalog0.35∼5 kHz∼14.7Resilience against the synaptic device fluctuations