Supervised Learning of Bayesian Network Parameters Made Easy

Supervised Learning of Bayesian Network Parameters Made Easy (2002)

Cached

Download Links

by Hannes Wettig , Peter Grünwald , Teemu Roos , Petri Myllymäki , Henry Tirri

Venue:	Level Perspective on Branch Architecture Performance, IEEE Micro-28
Citations:	4 - 1 self

Versions

Version History

Metadata Version 3

User correction supplied by SystemCorrections

Datum	Value	Source
TITLE	Supervised Learning of Bayesian Network Parameters Made Easy	user correction - Legacy Corrections
AUTHOR NAME	Hannes Wettig	user correction - Legacy Corrections
AUTHOR AFFIL	Computation Group; CWI	user correction - Legacy Corrections
AUTHOR ADDR	Helsinki Inst. for Inf. Tech. (HIIT) P.O. Box 94079	user correction - Legacy Corrections
AUTHOR NAME	Peter Grünwald	user correction - Legacy Corrections
AUTHOR NAME	Teemu Roos	user correction - Legacy Corrections
AUTHOR NAME	Petri Myllymäki	user correction - Legacy Corrections
AUTHOR AFFIL	University of Helsinki NL-1090 GB Amsterdam, The Netherlands.; & Helsinki University of Technology	user correction - Legacy Corrections
AUTHOR ADDR	P.O. Box 9800, FIN-02015 HUT, Finland	user correction - Legacy Corrections
AUTHOR NAME	Henry Tirri	user correction - Legacy Corrections
AUTHOR AFFIL	University of Helsinki NL-1090 GB Amsterdam, The Netherlands.; & Helsinki University of Technology	user correction - Legacy Corrections
AUTHOR ADDR	P.O. Box 9800, FIN-02015 HUT, Finland	user correction - Legacy Corrections
ABSTRACT	Bayesian network models are widely used for supervised prediction tasks such as classification. Usually the parameters of such models are determined using `unsupervised' methods such as maximization of the joint likelihood. In many cases, the reason is that it is not clear how to find the parameters maximizing the supervised (conditional) likelihood. We show how the supervised learning problem can be solved e#ciently for a large class of Bayesian network models, including the Naive Bayes (NB) and tree-augmented NB (TAN) classifiers. We do this by showing that under a certain general condition on the network structure, the supervised learning problem is exactly equivalent to logistic regression. Hitherto this was known only for Naive Bayes models. Since logistic regression models have a concave loglikelihood surface, the global maximum can be easily found by local optimization methods.	user correction - Legacy Corrections
YEAR	2002	user correction - Legacy Corrections
VENUE	Level Perspective on Branch Architecture Performance, IEEE Micro-28	INFERENCE
VENUE TYPE	CONFERENCE	INFERENCE
CITATIONS	13 found	ParsCit 1.0