Table 6.

Accuracy, Total # parameters and Energy Efficiency EE for STDP-trained SNN topologies (refer Table 7) of latency T = 100 and corresponding ANN on CIFAR10 data.

Model	ANN (T = 1)	SNN (T = 100)	#Parameters	EEConv/EEFull= $\frac{E_{ANN}(1\times )}{E_{SNN}}$
	(Accuracy%)
ResNet2	78.26	61.02	18.9 M	1.64x/1.16x
ResNet3	80.11	51.1	28.37 M	1.81x/1.28x
ResNet2x2	79.39	63.21	28.35 M	10.56x/1.78x

EE_Conv considers the energy calculated only for the convolutional/pooling layers excluding the FC layers, EE_Full considers the total energy of the network including the FC layers.