Question 4 5p Why is the rectifed linear unit (ReLU) a preferred choice of activation function for modern neural networks over the sigmoid activation function? Tick all the correct justifications. You should only tick an option if you're sure that it's correct. The ReLU activation can be computed more efficiently than a sigmoid activation, The RelU makes the neural network less prone to problem of vanishing gradients. The ReLU activation can be stored more efficiently than the sigmoid activation due to sparsity)