A Method for Integrating Expert Knowledge When Learning Bayesian Networks From Data

Abstract

Automatic learning of Bayesian networks from data is a challenging task, particularly when the data are scarce and the problem domain contains a high number of random variables. The introduction of expert knowledge is recognized as an excellent solution for reducing the inherent uncertainty of the models retrieved by automatic learning methods. Previous approaches to this problem based on Bayesian statistics introduce the expert knowledge by the elicitation of informative prior probability distributions of the graph structures. In this paper, we present a new methodology for integrating expert knowledge, based on Monte Carlo simulations and which avoids the costly elicitation of these prior distributions and only requests from the expert information about those direct probabilistic relationships between variables which cannot be reliably discerned with the help of the data.

Knowledge elicitation via sequential probabilistic inference for high-dimensional prediction

“…In addition to the differences coming from the learning task, one notable contrast between these works and our method is that their aim is to identify user preferences or opinions, whereas our goal is to use expert knowledge as an additional source of information for an improved prediction model, by integrating it with the knowledge coming from the (small n) data. As a probabilistic approach, our work relates to Cano et al (2011) and House et al (2015), where expert feedback is used for improved learning of Bayesian networks and for visual data exploration, respectively. In Sect.…”

“…We briefly describe two earlier works that can be seen as instances of it. Cano et al (2011) present a method for integrating expert knowledge into learning of Bayesian networks. The observation model is a multinomial Bayesian network with Dirichlet priors.…”

An empirical study of Bayesian network parameter learning with monotonic influence constraints

“…Cano et al [8] and Flores et al [35] proposed an interactive BN structure learning approach that iteratively queries the domain expert about the reliability of learnt edges. This interactive paradigm can be easily applied to help elicit edges' monotonic influences from experts and suggests that the theoretical method presented in this paper can be applied in practice (we do not use the interactive paradigm here because in our experiments we assume that any edges with monotonic influences are known and we simulate errors made by experts).…”

A Structure Learning Algorithm for Bayesian Network Using Prior Knowledge

“…To speed up the learning process, Corander et al proposed a method to learn structures of graphical model from training samples using the non-reversible parallel interacting MCMC-style computation [14] . Several heuristic algorithms also use prior knowledge to learn Bayesian network structure and obtain better results [15][16][17] . The algorithm proposed by Teyssier and Koller is based on the fact that the best structure is consistent with a node ordering, searched not over the space of structures but over the space orderings to reduce the search space [15] .…”

“…The algorithm proposed by Teyssier and Koller is based on the fact that the best structure is consistent with a node ordering, searched not over the space of structures but over the space orderings to reduce the search space [15] . Cano et al proposed an algorithm in which the domain experts are requested to submit their knowledge during the process and the system only asks the domain experts about the most uncertain structural feature and the presence or the absence of an edge [16] . de Campos and Castellano used a heuristic algorithm with statistic data and three types of prior knowledge: existence of arcs and/or edges, absence of arcs and/or edges, and ordering restrictions to learn the structure of Bayesian network [17] .…”