Explain how the hidden layer size influences the architecture of a feed$-$forward neural network. In doing so$,$ note what can happen if the hidden size is too large and what can happen if the hidden size is too small.