Minimum Error Tree Decomposition

This paper describes a generalization of previous meth­ ods for constructing tree-structured belief network with hidden variables. The major new feature of the de­ scribed method is the ability to produce a tree decom­ position even when there are errors in the correlation data among the input variables. This is an important extension of existing methods since the correlational co­ efficients usually cannot be measured with precision. The technique involves using a greedy search algorithm that locally minimizes an error function.