Extended tree augmented naive classifier

Cassio P. de Campos, Marco Cuccu, Giorgio Corani, Marco Zaffalon

Research output: Chapter in Book/Report/Conference proceedingChapterAcademicpeer-review

12 Citations (Scopus)


This work proposes an extended version of the well-known tree-augmented naive Bayes (TAN) classifier where the structure learning step is performed without requiring features to be connected to the class. Based on a modification of Edmonds’ algorithm, our structure learning procedure explores a superset of the structures that are considered by TAN, yet achieves global optimality of the learning score function in a very efficient way (quadratic in the number of features, the same complexity as learning TANs). A range of experiments show that we obtain models with better accuracy than TAN and comparable to the accuracy of the state-of-the-art classifier averaged one-dependence estimator.

Original languageEnglish
Title of host publicationProbabilistic Graphical Models: 7th European Workshop, PGM 2014, Utrecht, The Netherlands, September 17-19, 2014. Proceedings
EditorsLinda C. van der Gaag, Ad J. Feelders
Place of PublicationCham
Number of pages14
ISBN (Electronic)978-3-319-11433-0
ISBN (Print)978-3-319-11432-3
Publication statusPublished - 2014
Externally publishedYes
Event7th European Workshop Probabilistic Graphical Models (PGM 2014) - Utrecht, Netherlands
Duration: 17 Sept 201419 Sept 2014

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
ISSN (Print)0302-9743


Conference7th European Workshop Probabilistic Graphical Models (PGM 2014)
Abbreviated titlePGM 2014

Bibliographical note

(selected for special issue, blind peer reviewed by >3 reviewers)


Dive into the research topics of 'Extended tree augmented naive classifier'. Together they form a unique fingerprint.

Cite this