The vanishing gradient problem is a challenge for training deep neural networks using gradient$-$based learning methods. In this problem, you are asking basic questions about the vanishing gradient: $($i$)(2$ points$)$ Describe what the vanishing gradient problem is$.($ii$)(4$ points$)$ For the following four activation functions, which function has the vanishing gradient problem and which function doesn$$t$?$ Briefly discuss each activation function.$($iii$)(3$ points$)$ Consider a one$-$layer non$-$linear neural network y $=$ Sigmoid$($W x $+$ b$),$ if we have a large initialization of parameter W$,$ would this cause vanishing gradient or exploding gradient problem? Briefly discuss the reason.