In a nutshell, saying a function is Lipschitz means there exists a constant K such that the distance between two outputs is at most K times the distance between the inputs, and this K must work for all inputs in the domain. We look at the single variable case with an example, and then the multi variable case. In machine learning context, we want similar inputs to be classified as similar. Lipschitz functions provide a reliable way to do that.
37 Comments