Bijection required for inverse

Why is bijection required to get the inverse of a function?