Efficient Algorithms for Bayesian Network Parameter Learning from Incomplete Data

We propose a family of efficient algorithms for learning the parameters of a Bayesian network from incomplete data. Our approach is based on recent theoretical analyses of missing data problems, which utilize a graphical representa-tion, called the missingness graph. In the case of MCAR and MAR data, this graph need not be explicit, and yet we can still obtain closed-form, asymptotically consistent parameter esti-mates, without the need for inference. When this missingness graph is explicated (based on back-ground knowledge), even partially, we can obtain even more accurate estimates with less data. Em-pirically, we illustrate how we can learn the pa-rameters of large networks from large datasets, which are beyond the scope of algorithms like EM (which require inference).