Convergence of the graphical-model support approximation sequence

At least 3 years old · documented by

Let GG be an undirected graph with dd vertices. Let X⊆Rd\mathcal{X}\subseteq\mathbb{R}^d be an i.i.d. sample from a non-degenerate probability density f0f_0 on Rd\mathbb{R}^d. Let SG,X\mathcal{S}_{G,\mathcal{X}} be the support of the MLE f^\hat{f} over FG\mathcal{F}_G, and let DG(i)\mathcal{D}^{(i)}_G be the sets defined in the paper.

Support approximation conjecture. The sequence of sets converges to the MLE support:

lim⁡i→∞DG(i)=SG,X.\lim_{i\to\infty}\mathcal{D}^{(i)}_G=\mathcal{S}_{G,\mathcal{X}}.

For non-chordal graphs, computing this limit is described as an open problem, while numerical experiments for a four-cycle provide evidence for the asserted convergence of the sets and their volumes.

References

Primary source

Kaie Kubjas, Olga Kuznetsova, Elina Robeva, Pardis Semnani and Luca Sodomaco, “Log-concave density estimation in undirected graphical models”, arXiv:2206.05227 (2022).

Progress summary

Never refreshed

Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.

Solutions 0

No solutions have been posted yet.