It performs variable selection in a multivariate linear model by estimating the covariance matrix
of the residuals then use it to remove the dependence that may exist among the responses and eventually performs variable selection by using the Lasso criterion.
The method is described in the paper Perrot-Docks et al. (2017) .