$2$ CNN Layers

The first thing you'll need to do is implement some new layers. In particular we'll need:

ConvolutionalLayer Your constructor should take in a kernel width and height to intialize the

weights of the kernel accordingly. Note that for simplicity we will have a single kernel $($although you

may opt to support multiple ones$).$ This layer should have the methods listed below. For further

simplicity, we will not be passing the gradient back through this layer, so we don't need to put

anything in the backward or gradient methods.

forward

backward $-$ You can just put pass for this

gradient $-$ You can just put pass for this

updateWeights

crossCorrelate$2$D $-$ You will implement your own $2$D cross$-$correlation

MaxPoolLayer This just needs the forward, backward, and gradient methods.

FlatteningLayer This too just needs the forward, backward, and gradient methods.