CMStatistics 2018: Start Registration
View Submission - CMStatistics
B1702
Title: On integer linear programming approach to learning decomposable graphical models Authors:  Milan Studeny - Institute of Information Theory and Automation of the CAS (Czech Republic) [presenting]
Abstract: The decomposable graphical models, described by chordal undirected graphs, are crucial in famous local computation method, used widely in probabilistic graphical models. The basic ideas of the integer linear programming approach to learning these graphical models will be recalled. We propose to represent them by special zero-one vectors, which idea leads to the study of a special polytope, called chordal graph polytope. The focus will be on a conjecture; what are all facet-defining inequalities for this polytope.