I get the sense the algo is just a graph representation the dependency structure. Am wondering if there is anything written on learning the weights by optimizing some parameters rather than a forward calculation from the the correlation matrix.
$\begingroup$
$\endgroup$
