M. Studeny: On separation criterion and recovery algorithm for chain graphs. In Uncertainty in Artificial Intelligence. Proceedings of the 12th Conference (E. Horvitz, F. Jensen eds.), Morgan Kaufmann, San Francisco 1996, pp. 509-516.

The paper describes in a simple way some results of two (later published) papers:
M. Studeny: A recovery algorithm for chain graphs. International Journal of Approximate Reasoning 17 (1997), n. 2-3, pp. 265-293.
M. Studeny and R. R. Bouckaert: On chain graph models for description of conditional independence structures. The Annals of Statistics 26 (1998), n. 4, pp. 1434-1495.
and illustates them by examples.

A direct graphical separation criterion for chain graphs which generalizes the Pearl's d-separation criterion for Bayesian networks is introduced (recalled). It is equivalent to the classic moralization criterion for chain graphs and complete in the sense that for every chain graph there exists a strictly positive probability distribution satisfying exactly the independency statements derivable from the chain graph according to the separation criterion. Every class of Markov equivalent chain graphs can be uniquely described by a natural representative, called the largest chain graph. A recovery algorithm, which on basis of the (conditional) dependency model induced by a chain graph finds the corresponding largest chain graph, is presented.

AMS classification 68T30, 62H05

chain graph
Markov equivalence
largest chain graph
recovery algorithm

A pdf copy (converted postscript version) (1167kB) is available.