Backoff methods for estimating parameters of a Bayesian network
[edit]
Proceedings of The 3rd International Workshop on Advanced Methodologies for Bayesian Networks, PMLR 73:33, 2017.
Abstract
Various authors have highlighted inadequacies of BDeu type scores and this problem is
shared in parameter estimation. Basically, Laplace estimates work poorly, at least because
setting the prior concentration is challenging. In 1997, Freidman et al suggested a simple
backoff approach for Bayesian network classifiers (BNCs). Backoff methods dominate in
in ngram language models, with modified KneserNey smoothing, being the best known,
and a Bayesian variant exists in the form of PitmanYor process language models from
Teh in 2006. In this talk we will present some results on using backoff methods for Bayes
network classifiers and Bayesian networks generally. For BNCs at least, the improvements
are dramatic and alleviate some of the issues of choosing too dense a network.
Related Material


