Stochastic Algorithms for Searching Causal Orderings in Bayesian Networks
Luis M. de Campos, Juan F. Huete · Studies in fuzziness and soft computing · 2002
The structure of a Bayesian network depends heavily on the ordering of its variables: given any ordering it is always possible to build a Bayesian network whose arcs are consistent with the initial ordering; however, the topology of the network, and therefore the number of conditional independence relationships that may be explicitly represented can vary greatly from one ordering to another. As a sparse representation is always preferable to a denser representation of the same model, the task of determining the ordering giving rise to the network with minimum number of arcs is important. In this work we propose methods to obtain a good approximation to the optimal ordering, using only partial information. More precisely, we only use conditional independence relationships of order zero and one, and search for the ordering which best preserves this information. The search process will be guided by genetic algorithms and simulated annealing. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.