Learning Selectively Conditioned Forest Structures with Applications to DNBs and Classification
journal contributionposted on 2007-01-01, 00:00 authored by Brian D. Ziebart, Anind K. Dey, J. Andrew Bagnell
Dealing with uncertainty in Bayesian Net- work structures using maximum a posteriori (MAP) estimation or Bayesian Model Averaging (BMA) is often intractable due to the superexponential number of possible directed, acyclic graphs. When the prior is decomposable, two classes of graphs where eﬃcient learning can take place are tree- structures, and ﬁxed-orderings with limited in-degree. We show how MAP estimates and BMA for selectively conditioned forests (SCF), a combination of these two classes, can be computed eﬃciently for ordered sets of variables. We apply SCFs to temporal data to learn Dynamic Bayesian Networks having an intra-timestep forest and inter-timestep limited in-degree structure, improving model accuracy over DBNs without the combination of structures. We also apply SCFs to Bayes Net classiﬁcation to learn selective forest- augmented Na¨ıve Bayes classiﬁers. We argue that the built-in feature selection of selective augmented Bayes classiﬁers makes them preferable to similar non-selective classiﬁers based on empirical evidence.