I am reading about K-means algorithm, and trying to explain it to myself in one sentence. However, I am a bit confused. I have came up with following definitions and I am not sure whether which one is more precise to describe it. Which definition makes more sense? Also, kernel of K-means can be only arbitrary or Gaussian? I know that it is used in Gaussian mixture models, but I am not sure.
- nonconvex algorithm to cluster the data
- kernelized version of means algorithm where the kernel is arbitrary
- kernelized version of means algorithm where the kernel is Gaussian
