Consider the following DNN for image classification of $10$ classes. The dataset consists of colour images of size $16$x$16.[2+2\mathrm{.}5+1\mathrm{.}5=6$ marks$]$ Input Layer inputs $=$ keras. Input $($shape$=($# #A##$))$ Layer $1$ x $=$ layers. Dense $(30,$ activation$=$'relu', use$\_$bias$=$False$)($inputs$)$ #Layer $2$ x $=$ layers. Dense $(60,$ activation$=$'relu'$)($x$)$ model $=$ keras. Model $($inputs$=$inputs, outputs$=$outputs, name$=$"model"$)$ model.compile $($optimizer $=$ 'adam', loss $=$ #D##$,$ metrics$=[\text{'}$accuracy$\text{'}])$ A$.$ Write the value or code of ##A## $=$B## ##C## and ##D# to complete the network. B$.$ If we add a dropout layer of value $0\mathrm{.}4$ after layer $2,$ I. What will be the total number of active neurons in layer $2$ during a$)$ model training b$)$ model testing? II$.$ What will be the total number of parameters in the network? What is the change in the number of parameters compared to the original network? C$.$ Assume that the image dataset consists of $5000$ train images and $1000$ test images. We run for $10$ epochs with batch size $=32,$ learning rate $($LR$)=0\mathrm{.}01$ and SGD optimizer. During model training, I. What will be the size of the batch in the last iteration? II$.$ What is the number of times we perform forward propagation? III. What is the number of times we perform backpropagation? And why?