Huh–Sturmfels conjectural inductive formula for maximum likelihood degree

Let XX be an irreducible algebraic subvariety of (C)n({\mathbb C}^*)^n contained in

H={p1++pn=1},H=\{p_1+\cdots+p_n=1\},

and let Xˉ\bar X be its closure in (C)n1×C({\mathbb C}^*)^{n-1}\times{\mathbb C}. Suppose that Xˉ{pn=0}\bar X\cap\{p_n=0\} is nonempty, reduced, and irreducible. The quantity MLdeg(Xˉ{pn=0})\operatorname{MLdeg}(\bar X\cap\{p_n=0\}) is its maximum likelihood degree as a subvariety of (C)n1×{pn=0}({\mathbb C}^*)^{n-1}\times\{p_n=0\}, and MLdeg(Xλn=0)\operatorname{MLdeg}(X|_{\lambda_n=0}) counts the points of XregX_{\operatorname{reg}} where

λ1dp1p1++λn1dpn1pn1\frac{\lambda_1dp_1}{p_1}+\cdots+\frac{\lambda_{n-1}dp_{n-1}}{p_{n-1}}

degenerates for generic (λ1,,λn1)Cn1(\lambda_1,\ldots,\lambda_{n-1})\in{\mathbb C}^{n-1}. Huh–Sturmfels' conjectural inductive formula.

MLdeg(X)=MLdeg(Xˉ{pn=0})+MLdeg(Xλn=0).\operatorname{MLdeg}(X)=\operatorname{MLdeg}(\bar X\cap\{p_n=0\})+\operatorname{MLdeg}(X|_{\lambda_n=0}).

The formula is intended to give an inductive computation of the maximum likelihood degree by passing to the boundary pn=0p_n=0 and restricting the likelihood parameters; the source does not specify its resolution status.

Sources & referencesView supporting material

Primary source

Nero Budur and Botong Wang, “Bounding the maximum likelihood degree”, arXiv:1411.3486 (2015).

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.