Question $5(15$ points$)$

Instead of the ReLU which we saw as a widely used activation fuction in DLNs $($Deep Learning Networks$)$ an alternative to ReLU called Exponential LinearUnit $($ELU$)$

ELU$(x)=\{\begin{array}{cc}x,& x0\\ (e^x-1),& x<mn\; id="EditorElement\_827931">0</mn>\end{array}$

is recommended by a friend of yours. Does ELU satisfy the requirements of an activation function $?$