16.04.2019 um 16:15 Uhr in 69/125:
Alexandros Grosdos Koutsoumpelias (Universität Osnabrück)
Combinatorial methods for algebraic problems in graphical models
Graphical models encode causal relations among random variables and find a range of applications in statistics. In the case of Gaussian random variables, the covariance matrix associated to the graph consists of rational functions in a polynomial ring. Goal of this talk is to shed light on algebraic questions related to the covariance matrix, such as finding its ideal or studying the fibers of the corresponding map. We present three combinatorial tools to do this, namely treks on graphs, graph decomposition and nested determinants.