Multivariate Edgeworth-Based Entropy Estimation
Marc M. Van Hulle · 2006
We develop the general, multivariate case of the Edgeworth approximation of differential entropy, and introduce an approximate formula for Gaussian mixture densities. We use these entropy approximations in a new algorithm for selecting the optimal number of clusters in a data set, and in a new mutual information test with which one can statistically decide whether a distribution can be factorized along a given set of axes