Table 1. VGG configuration for CIFAR10/100 (Simonyan & Zisserman, 2015).

Layer	Output size	Network configuration
conv-1	$32\times 32$	$\left[3\times 3,64\right]\times 2$
	$16\times 16$	maxpool
conv-2	$16\times 16$	$\left[3\times 3,64\right]\times 2$
	$8\times 8$	maxpool
conv-3	$8\times 8$	$\left[3\times 3,128\right]\times 2$
	$4\times 4$	maxpool
	2 × 2	AdaptiveAvgPool
	1 × 1	FC-10/100

Notes:

$[m\times m,n]\times k$, $m\times m$: convolution kernel size, n: output channels; k: the repeat number of the layer.

Each convolutional layer is followed by a BN and ReLU layer.

maxpool: [ $2\times 2$] kernel with stride 2.

Output size: $width\times height$.