Why would one use zero$-$padding in CNNs$?$ How much does it contribute to the number of parameters in the network? Group of answer choices: $1)$ Zero$-$padding is applied to reduce overfitting by adding extra data to the input. It moderately increases the number of parameters, as each padding value introduces new weights. $2)$ Zero$-$padding is used to decrease the computational cost of convolutions by reducing the effective size of the input. It contributes to the parameter count by adding extra layers that depend on the size of the padding. $3)$ Zero$-$padding is used to maintain the spatial dimensions of the input after convolution, ensuring that the output size remains consistent with the input size. It does not contribute to the number of parameters, as padding only affects the input, not the learnable weights. $4)$ Zero$-$padding is used to increase the number of parameters in the network, as it adds additional learnable weights. It significantly contributes to the parameter count.