R/penalty.as.matrix.R

In JGL: Performs the Joint Graphical Lasso for Sparse Inverse Covariance Estimation on Multiple Classes

penalty.as.matrix <-
function(lambda,p,penalize.diagonal)
{
	# for matrix penalties:  check dim and symmetry:
	if(is.matrix(lambda))
	{
		if(sum(lambda!= t(lambda))>0) {stop("error: penalty matrix is not symmetric")}
		if(sum(abs(dim(lambda)-p))!=0 ) {stop("error: penalty matrix has wrong dimension")}
	}
	# for scalar penalties: convert to matrix form:
	if(length(lambda)==1) {lambda=matrix(lambda,p,p)}
	# apply the penalize.diagonal argument:
	if(!penalize.diagonal) {diag(lambda)=0}
	return(lambda)
}