Theory Refinement on Bayesian Networks

By Wray L. Buntine. 1991

Theory refinement is the task of updating a domain theory in the light of new cases, to be done automatically or with some expert assistance. The problem of theory refinement under uncertainty is reviewed here in the context of Bayesian statistics, a theory of belief revision. The problem is reduced to an incremental learning task as follows: the learning system is initially primed with a partial theory supplied by a domain expert, and thereafter maintains its own internal representation of alternative theories which is able to be interrogated by the domain expert and able to be incrementally refined from data. Algorithms for refinement of Bayesian networks are presented to illustrate what is meant by "partial theory", "alternative theory representation", etc. The algorithms are an incremental variant of batch learning algorithms from the literature so can work well in batch and incremental mode.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Causal Inference through a Witness Protection Program (2014, methodology): “To decide whether the premises in Rule 1 hold, we merely perform Bayesian model selection with the BDeu score (Buntine, 1991) between the full graph {W → X,W → Y,X → Y } (conditional on Z) and the graph with the edge W → Y removed.”
Learning Bayesian networks: The combination of knowledge and statistical data (1994, methodology): “More recently, AI researchers and statisticians have begun to investigate methods for learning Bayesian networks, including Bayesian methods (Cooper & Herskovits, 1991; Buntine, 1991; Spiegelhalter et al., 1993; Dawid & Lauritzen, 1993; Heckerman et al., 1994), quasi-Bayesian methods (Lam &”
Exposing the probabilistic causal structure of discrimination (2015, background): “We start byobserving that,while reconstructing aBayesian Network is a known NP-hard problem, when an ordering among the variables in the network is given, finding the maximum likelihood network is not NP-hard anymore [3,4].”
Incremental learning of privacy-preserving Bayesian networks (2013, methodology): “In [3] a generalized approach of K2 algorithm is first proposed, and then some guidelines are presented to convert that algorithm into an incremental approach.”
Learning the Structure of Deep Sparse Graphical Models (2009, background): “…[2006], Elidan et al. [2000], Friedman [1998], Ramachandran and Mooney [1998]), most previous work on learning the structure of belief networks has focused on the case where all units are observed [Buntine, 1991, Friedman and Koller, 2003, Heckerman et al., 1995, Koivisto and Sood, 2004].”
Regime Aware Learning (2016, background): “Refining or updating the structure and conditional distributions of a BN in response to new data has been studied for some time (Buntine, 1991; Lam and Bacchus, 1994; Friedman and Goldszmidt, 1997; Lam, 1998).”
Efficient Structure Learning of Bayesian Networks using Constraints (2011, methodology): “1 In this paper, we consider some well-known score functions: the Bayesian Information Criterion (BIC) (Schwarz, 1978) (which is equivalent to the Minimum Description Length), the Akaike Information Criterion (AIC) (Akaike, 1974), and the Bayesian Dirichlet (BD) (Cooper and Herskovits, 1992), which ”
Ordering-Based Search: A Simple and Effective Algorithm for Learning Bayesian Networks (2005, methodology): “Our approach is based on the fundamental observation [Buntine, 1991; Cooper and Herskovits, 1992] that, given an ordering (cid:30) on the variables in the network, (cid:12)nding the highest-scoring network consistent with (cid:30) is not NP-hard.”
On Inclusion-Driven Learning of Bayesian Networks (2003, unknown): “Throughout this paper we have used the BDeu scoring metric, which corresponds to the BDe metric from Heckerman et al. (1995) with uniform hyper-Dirichlet priors (Buntine, 1991).”
Exact Bayesian Structure Discovery in Bayesian Networks (2004, methodology): “Since the seminal works of Buntine (1991) and Cooper and Herskovits (1992) numerous algorithms have been presented for these structure learning tasks.”
Online Machine Learning in Big Data Streams (2018, methodology): “Bayesian networks were one of the earliest applications for online learning [88], with the first methods mostly using mini-batch updates [38].”
A Tutorial on Learning with Bayesian Networks (1999, background): “Several authors have discussed such assumptions and corresponding methods for deriving priors (Cooper and Herskovits, 1991, 1992; Buntine, 1991; Spiegelhalter et al., 1993; Heckerman et al., 1995b; Heckerman and Geiger, 1996).”
Being Bayesian about Network Structure (2000, background): “(6) was first mentioned by Buntine [2], but the ramifications for Bayesian model averaging were not pursued.”
Learning Bayesian Networks with Local Structure (1996, methodology): “Another popular scoring metric for learning Bayesian networks is the Bayesian based metric described by Heckerman, Geiger and Chickering (HGC) [Heckerman, Geiger, and Chickering 1995] (which is based on earlier work of [Cooper and Herskovits 1992; Buntine 1991b]).”
Contrast Pattern-Based Classification for Bot Detection on Twitter (2019, background): “C. STATE-OF-THE-ART CLASSIFIERS For comparing the classification results obtained by our proposal regarding other state-of-the-art classifiers, we have selected the following classifiers: BN:Bayesian Network [63] is a directed acyclic graph that models probabilistic relationships among a set of rand”
Advances in Exact Bayesian Structure Discovery in Bayesian Networks (2006, methodology): “This paper puts forward the Bayesian approach to structure learning (Buntine, 1991, Cooper and Herskovits, 1992, Friedman and Koller, 2003). We improve a recently found exact algorithm of Koivisto and Sood (2004), which computes the marginal posterior probability of any given subnetwork, e.”