In a neural network, the sigmoid function is used to add nonlinearity. A bounded sigmoid function is the hyperbolic tangent function, or this translated to the range (0,1); an unbounded sigmoid function is the inverse hyperbolic cosine function.
See also: Sigmoid function