Work

Completions to discrete probability distributions in log-linear models

Year: 2024

Type: article

Abstract: Completion problems, of recovering a point from a set of observed coordinates, are abundant in applications to image reconstruction, phylogenetics, and data science. We consider a completion problem coming from algebraic statistics: to describe the completions of a point to a probability distribution lying in a given log-linear model. When there are finitely many completions, we show that these points either have a unique completion or two completions to the log-linear model depending on the set of observed coordinates. We additionally describe the region of points which have a completion to the log-linear model.

Source: Algebraic Statistics

DOI: 10.2140/astat.2024.15.225

Authors: May Cai, Cecilie Olesen Recke, Thomas Yahl

Institutions: Georgia Institute of Technology, University of Copenhagen, University of Wisconsin–Madison

Language: English

Topic: Polynomial and algebraic computation

Subfield: Computational Theory and Mathematics

Field: Computer Science

Domain: Physical Sciences

Open Access status: diamond

Funders: National Science Foundation, Novo Nordisk, Novo Nordisk Fonden

Awards: Institute for Mathematical and Statistical Innovation, G3211459053, Polytopes and Real Tropical Geometry