Q$)(50$ Marks$)$

Deep learning consists of ConNN layers followed by a pooling layer and MLP NN$.$ Such networks mainly are applied to find the features of multimap input images. Such levels are different; low, medium, and high levels.

Construct a network that deals with the three levels. For mo input maps of layer $(6-1)$ assign the cubic filters that are used to get $D$ maps at layer $(6).$ Take VGG$16$ as an example.

How to get the layer $(6)$ output for $m$ maps of layer $(b-1).$ Take VGG$16$ as an example to demonstrate the mathematical analysis.

To learn layer $(6-1)$ you need to update the weights of layer $(6),$ how to achieve that considering that any map at layer $(6)$ is due to the sum effect of all maps at layer $(b-1).$