Please cite one or more of the following relevant papers if you use this code.
Code for Square Root Graphical Model (SQR) is based on:
David I. Inouye, Pradeep Ravikumar, Inderjit S. Dhillon. Square Root Graphical Models: Multivariate Generalizations of Univariate Exponential Families that Permit Positive Dependencies International Conference on Machine Learning (ICML), 2016. https://www.davidinouye.com/publication/inouye-2016-square/inouye-2016-square.pdf
Code and data are also provided for the following review paper.
David I. Inouye, Eunho Yang, Genevera I. Allen, Pradeep Ravikumar.
A review of multivariate distributions for count data derived from the Poisson distribution.
Wiley Interdisciplinary Reviews (WIREs): Computational Statistics, 9:3, 2017. doi: 10.1002/wics.1398
arXiv preprint: https://arxiv.org/pdf/1609.00066.pdf
The implementation of Square Root Graphical Models for the Poisson distribution is based on the following arXiv paper:
David I. Inouye, Pradeep Ravikumar, Inderjit S. Dhillon
Generalized Root Models: Beyond Pairwise Graphical Models for Univariate Exponential Families
arXiv preprint arXiv:1606.00813, 2016.
arXiv preprint: https://arxiv.org/pdf/1606.00813.pdf
You must install the R packages VineCopula and XMRF for the vine copula and TPGM models to work respectively.
The main demo file is demo_comparison.m but the demo_comparison_check.m file checks that all the methods run to completion for a really small dataset.
The 6 datasets used in the paper are included as simple MAT files in the data folder.