The Johnson–Lindenstrauss lemma states that a set of points in a high-dimensional space can be embedded into a space of much lower dimension in such a way that distances between the points are nearly preserved. To do so, it requires random variables to be sampled from a normal distribution. In this proof, it states that the normal distribution is used because a normalized vector of Gaussian variables is uniformly distributed on the surface of a d-dimensional sphere. The proof of this can be found here.

Do any other distributions have this same property such that they can be used in the dimensionality reduction?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.