Table 3.

Network structure.

Dataset	Network structure
DVS128 Gesture	Input-128C3-MP2-128C3-MP2-128C3- MP2-128C3-MP2-128C3-MP2-0.5DP-512FC-0.5DP-110FC-Voting-11
CIFAR10-DVS	Input-64C3-128C3-AP2-256C3-256C3- AP2-512C3-512C3-AP2-512C3-512C3- AP2-100FC-Voting-10
SHD	Input-128FC-128FC-100FC-Voting-20
N-MNIST	Input-128C3-AP2-128C3-AP2-0.5DP- 2048FC-0.5DP-100FC-Voting-10

xCy/MPy/APy denotes the Conv2D/MaxPooling/AvgPooling layer with output channel = x and kernel size = y×y, nFC denotes the fully connected layer with output feature = n, mDP is the spiking dropout layer with dropout ratio m. The spiking neurons are added behind all xCy and nFC. The STSC modules are inserted before all xCy on DVS128 Gesture, CIFAR10-DVS, and N-MNIST; while inserted before all nFC on SHD.