Currently, neural networks typically use an activation function characterized by:

A.a function with a default 0 value and returns x only if the input x is positive.

B.None of the other options

C.a logit function

D.a hinge loss function

E.a logistic function