Univariant Approximation by Superpositions of a Sigmoidal Function

Every function of bounded variation can be approximated by a linear combination of a fixed sigmoidal function of n terms to an error 0(1/n) in the uniform norm.