Question $7$

Consider the following convolutional neural network that has two convolutional $($CONV$)$ layers, one max pooling $($POOL$)$ layers, and two fully$-$connected $($FC$)$ layers. $n$@$wh$ means the feature map has $n$ channels and the resolution is $w$ by $h\mathrm{.}1m$ means there are $m$ elements in the output vector of the FC laver. $($Total: $20$ marks$)$

a$)$ Assume there is no zero padding for CONV$1$ and CONV$2.$ Find the stride, filter size, the numbers of input channels and output channels of the two convolutional layers $($Note: the filter size should be smaller than $1010).(12$ marks$)$

b$)$ Find the numbers of parameters for the two convolutional layers and the two fully$-$connected layers $($do not consider bias$).(8$ marks$)$