Compressible Distributions for High-dimensional Statistics

Rémi Gribonval, Volkan Cevher, Mike E. Davies

We develop a principled way of identifying probability distributions whose independent and identically distributed (iid) realizations are compressible, i.e., can be well-approximat…