Log-concave density estimation in undirected graphical models

06/10/2022
by   Kaie Kubjas, et al.
0

We study the problem of maximum likelihood estimation of densities that are log-concave and lie in the graphical model corresponding to a given undirected graph G. We show that the maximum likelihood estimate (MLE) is the product of the exponentials of several tent functions, one for each maximal clique of G. While the set of log-concave densities in a graphical model is infinite-dimensional, our results imply that the MLE can be found by solving a finite-dimensional convex optimization problem. We provide an implementation and a few examples. Furthermore, we show that the MLE exists and is unique with probability 1 as long as the number of sample points is larger than the size of the largest clique of G when G is chordal. We show that the MLE is consistent when the graph G is a disjoint union of cliques. Finally, we discuss the conditions under which a log-concave density in the graphical model of G has a log-concave factorization according to G.

READ FULL TEXT

Please sign up or login with your details

Forgot password? Click here to reset