# Why is the softmax used to represent a probability distribution?

In the machine learning literature, to represent a probability distribution, the softmax function is often used. Is there a reason for this? Why isn't another function used?