Employing Maximum Mutual Information for Bayesian Classification

Gerven, Marcel van; Lucas, Peter

doi:10.1007/978-3-540-30547-7_20

Cited by 3 publications

(1 citation statement)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Information theory has been applied to BNs for the purpose of improving the training process and for performance analysis [14,28,44,45]. For instance, the cross-entropy measure between a trained BN and a known underlying probability distribution is utilized in [45] for evaluating and comparing the learning performance of several training algorithms.…”

Section: Development Of Bayesian Network Performance Metricsmentioning

confidence: 99%

Constructing Bayesian networks for criminal profiling from limited data

Baumgartner

Ferrari

Palermo

2008

Knowledge-Based Systems

View full text Add to dashboard Cite

Section: Development Of Bayesian Network Performance Metricsmentioning

confidence: 99%

Constructing Bayesian networks for criminal profiling from limited data

Baumgartner

Ferrari

Palermo

2008

Knowledge-Based Systems

View full text Add to dashboard Cite

Using Background Knowledge to Construct Bayesian Classifiers for Data-Poor Domains

Gerven¹,

Lucas²

2005

Research and Development in Intelligent Systems XXI

Self Cite

View full text Add to dashboard Cite

The development of Bayesian classifiers is frequently accomplished by means of algorithms which are highly data-driven. Often, however, sufficient data are not available, which may be compensated for by eliciting background knowledge from experts. This paper explores the trade-offs between modelling using background knowledge from domain experts and machine learning using a small clinical dataset in the context of Bayesian classifiers. We utilised background knowledge to improve Bayesian classifier performance, both in terms of classification accuracy and in terms of modelling the structure of the underlying joint probability distribution. Relative differences between models of differing structural complexity, which were learnt using varying amounts of background knowledge, are explored. It is shown that the use of partial background knowledge may significantly improve the quality of the resulting classifiers.

show abstract

Discrete Bayesian Network Classifiers

2014

View full text Add to dashboard Cite

We have had to wait over 30 years since the naive Bayes model was first introduced in 1960 for the so-called Bayesian network classifiers to resurge. Based on Bayesian networks, these classifiers have many strengths, like model interpretability, accommodation to complex data and classification problem settings, existence of efficient algorithms for learning and classification tasks, and successful applicability in real-world problems. In this article, we survey the whole set of discrete Bayesian network classifiers devised to date, organized in increasing order of structure complexity: naive Bayes, selective naive Bayes, seminaive Bayes, one-dependence Bayesian classifiers, k -dependence Bayesian classifiers, Bayesian network-augmented naive Bayes, Markov blanket-based Bayesian classifier, unrestricted Bayesian classifiers, and Bayesian multinets. Issues of feature subset selection and generative and discriminative structure and parameter learning are also covered.

show abstract

Employing Maximum Mutual Information for Bayesian Classification

Cited by 3 publications

References 9 publications

Constructing Bayesian networks for criminal profiling from limited data

Constructing Bayesian networks for criminal profiling from limited data

Using Background Knowledge to Construct Bayesian Classifiers for Data-Poor Domains

Discrete Bayesian Network Classifiers

Contact Info

Product

Resources

About