nodag

structure learning without imposing acyclicity

This repository contains fortran implementation of a proximal gradient method to estimate the structure of a structural equation model (SEM) without imposing the acyclicity constraint. See the preprint on arXiv:2006.03005.

The subroutine NODAG solves the following l1-penalized minus log-likelihood minimization:

minimize  -2log(det(A)) + trace( Sigma AA^t) + lambda ||A||_1 
A invertible 

use

The fortran subroutine NODAG can be easily used both from python and R

• python: compile nodag.f with f2py using f2py -llapack -c -m nodag nodag.f
• R: compile nodag.f with R CMD SHLIB nodag.f -llapack -lblas

Check the provided examples to see how to load and call the subroutine.

versions

• 0.0.3 July 6, 2020
• 0.0.2 June 12, 2020
• 0.0.1 June 6, 2020