logistic function

A logistic function is a form of sigmoid function often used as a weak threshold in an neural network. Like a step function it maps unbounded values to a finite [0,1] range and is zero for large negative values and one for large postive values. However, unlike a step function the progress is smooth with a near linear section in the middle and gently asymptotes for more extreme values. This is important for machine learning as, in general, it is easier to learn continuous features.