HIGH-DIMENSIONAL GRAPHICAL MODELLING USING TREES AND FORESTS
David Edwards · 2010
The seminar will describe some methods allowing the application of mixed graphical models (hybrid Markov networks) to high-dimensional datasets. These are based on (i) various extensions to the Chow-Liu algorithm and (ii) a greedy algorithm to find the strongly decomposable model with minimum AIC/BIC. The former is highly efficient and the latter less so. The methods are implemented in an R library, gRapHD, and will be illustrated using some biological data set.