Benchmark Network I

Input type: Poisson spikes (NE15-Poissonian)

Network:

• One input layer network

• Fully connected decision layer(50x10)

• Current based LIF-exp neurons:

  Parameter	Values	Units
  C_m	0.25	nF
  Tau_m	20	ms
  Tau_refrac	2	ms
  V_reset	-70	mV
  V_rest	-65	mV
  V_thresh	-50	mV

Training:

• K-means clusters as preprocessing (to select appropriate decision neuron)
• Supervised STDP
• A total of 18,000 biological seconds

Testing:

• 5k Hz input rate
• 1s per test and 200ms interval silence (in biological time)

Performance:

• 92.99% accuracy
• 13.82ms latency
• 4.17M Sopbs

Hardware Platform	Accuracy (%)	Sim Time (s)	Enery (KJ)	Ref
SpiNNaker	92.99	12,000	4.92