We describe a graphical representation of probabilistic relationships --- an alternative to the Bayesian network --- called a dependency network. Like a Bayesian network, a dependency network has a graph and probability component. The graph component is a (cyclic) directed graph such that a node's parents render that node independent of all other nodes in the network. The probability component consists of the probability of a node given its parents for each node (as in a Bayesian network). We identify several basic properties of this representation, and describe its use in density estimation, collaborative filtering (the task of predicting preferences), and the visualization of predictive relationships.

}, author = {D. Heckerman and D.M. Chickering and C. Meek and R. Rounthwaite and C. Kadie}, institution = {Microsoft Research}, journal = {Journal of Machine Learning Research}, month = {October}, number = {MSR-TR-2000-16}, pages = {49-75}, publisher = {Journal of Machine Learning Research}, title = {Dependency Networks for Inference, Collaborative Filtering, and Data Visualization}, url = {http://research.microsoft.com/apps/pubs/default.aspx?id=64334}, volume = {1}, year = {2000}, }