Extended Bayesian Classifiers

For some years, I have been intrigued with the naive Bayesian classifier, an algorithm for supervised learning that stores a single probabilistic summary for each class and that assumes conditional independence of the attributes given the class. Despite these simpiifying assumptions, in many domains naive Bayes gives results as good or better than much more sophisticated approaches to induction.

In addition to analyses of naive Bayes' behavior, my colleagues (Stephanie Sage and George John) and I have extended the basic algorithm along a number of fronts, but striving to keep the representational ability (and thus the inductive bias) within reasonable bounds. This constrasts with much of the recent work on learning with unrestricted Bayesian networks. Our studies have led to the publications shown below.


Related Publications

Langley, P., & Sage, S. (1999). Tractable average-case analysis of naive Bayesian classifiers. Proceedings of the Sixteenth International Conference on Machine Learning (pp. 220-228). Bled, Slovenia: Morgan Kaufmann.

John, G. H., & Langley, P. (1995). Estimating continuous distributions in Bayesian classifiers. Proceedings of the Eleventh Conference on Uncertainty in Artificial Intelligence (pp. 338-345). Montreal, Quebec: Morgan Kaufmann.

Langley, P., & Sage, S. (1994). Induction of selective Bayesian classifiers. Proceedings of the Tenth Conference on Uncertainty in Artificial Intelligence (pp. 399-406). Seattle, WA: Morgan Kaufmann.

Langley, P. (1993). Induction of recursive Bayesian classifiers. Proceedings of the 1993 European Conference on Machine Learning (pp. 153-164). Vienna: Springer-Verlag.

Langley, P., Iba, W., & Thompson, K. (1992). An analysis of Bayesian classifiers. Proceedings of the Tenth National Conference on Artificial Intelligence (pp. 223-228). San Jose, CA: AAAI Press.

For more information, send electronic mail to langley@isle.org


© 1997 Institute for the Study of Learning and Expertise. All rights reserved.