Identification of Latent Variables From Graphical Model Residuals

This is a work in progress where we are developing an algorithm to recover the causal structure among observed variables in the presence of a latent confounder (an unobserved variable that drives multiple observed variables) that drives a large number of observed variables. An example of the latent confounders that we are interested in recovering are so-called master regulators which drives many genes on a gene network.