Learning Mixtures of DAG Models
By Bo Thiesson, Christopher Meek, David Maxwell Chickering, David E. Heckerman. 1998
We describe computationally efficient methods for learning mixtures in which each component is a directed acyclic graphical model (mixtures of DAGs or MDAGs). We argue that simple search-and-score algorithms are infeasible for a variety of problems, and introduce a feasible approach in which parameter and structure search is interleaved and expected data is treated as real data. Our approach can be viewed as a combination of (1) the Cheeseman-Stutz asymptotic approximation for model posterior probability and (2) the Expectation-Maximization algorithm. We evaluate our procedure for selecting among MDAGs on synthetic and real examples.
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
- Rónán Daly
- Qiang Shen
- Stuart Aitken
- Kevin P. Murphy
- Daniel Eaton
- Yoshua Bengio
- Liang-Chieh Chen
- Antonio Torralba
Citations in the corpus
- Modelling Gene Expression Data using Dynamic Bayesian Networks (2006, background): “This has been done for Gaussian networks [TMCH98] and discrete, undirected trees2 [MJM98]; unfortunately, there are severe computational difficulties in generalizing this technique to arbitrary, discrete networks.”
- Belief net structure learning from uncertain interventions (2007, background): “If all the backbone nodes depend on Ii, then we have a mixture of belief networks (sometimes called a “Bayesian multinet” (Thiesson et al., 1998; Bilmes, 2000)); but if only some nodes depend on Ii, we get a “factored mixture”, in which only some components of the density depend on the mixing indica”
- Learning Bayesian networks: approaches and issues (2011, background): “Thiesson et al. (1998a, 1998b) describe an approach to learning what they call mixtures of Bayesian networks and mixtures of DAGs.”