The Kolmogorov-Arnold representation theorem says any function witj N inputs can be created using a two layer network of at most 2 N + 1 nodes where each noed's output is a nin-linear function of the sum of its imputs. However, the non-linear functions at each node may al be different and very complex including discontinuities which are effectuvely, meaning that the result at first appears to only have very abstract thoeretical interest. However, Kolmogorov-Arnold Networks are showing promise at being able to create usable neural networks using this result.
Defined on page 146
Used on page 146
Links:
Wikipedia: Kolmogorov–Arnold representation theorem