Which of the following methods can help improve the performance of a Convolutional Neural Network?

Options:

$1)$ Using U regularization on the weights

$2)$ Data augmentation techniques film rotation, scaling, and flipping

$3)$ Applying transfer learning from a pre$-$trained model

$4)$ Increasing the depth of the network with more convolutional layers