Structured taxonomies characterize several real world prob- lems, ranging from text categorization, to video annotation and protein function prediction. In this context “flat” learning methods may intro- duce inconsistent predictions, while structured output-aware learning methods can improve the accuracy of the predictions by exploiting the hierarchical relationships between classes. We propose a novel hierarchi- cal ensemble method able to provide theoretically guaranteed consistent predictions for any Directed Acyclic Graph (DAG)-structured taxonomy, and consequently also for any taxonomy structured according to a tree. Results with a complex real-world DAG-structured taxonomy involving about one thousand classes and twenty thousand of examples show that the proposed hierarchical ensemble approach significantly improves flat methods, especially in terms of precision/recall curves.
A Hierarchical Ensemble Method for DAG-Structured Taxonomies / P.N. Robinson, M. Frasca, S. Köhler, M. Notaro, M. Re, G. Valentini (LECTURE NOTES IN COMPUTER SCIENCE). - In: Multiple Classifier Systems : 12th International Workshop, MCS 2015, Günzburg, Germany, June 29 - July 1, 2015, Proceedings / [a cura di] F. Schwenker, F. Roli, J. Kittler. - Berlin : Springer, 2015 Jun. - ISBN 9783319202471. - pp. 15-26 (( Intervento presentato al 12. convegno MCS (Multiple Classifier Systems) International Workshop tenutosi a Gunzburg nel 2015 [10.1007/978-3-319-20248-8_2].
A Hierarchical Ensemble Method for DAG-Structured Taxonomies
M. FrascaSecondo
;M. Notaro;M. RePenultimo
;G. ValentiniUltimo
2015
Abstract
Structured taxonomies characterize several real world prob- lems, ranging from text categorization, to video annotation and protein function prediction. In this context “flat” learning methods may intro- duce inconsistent predictions, while structured output-aware learning methods can improve the accuracy of the predictions by exploiting the hierarchical relationships between classes. We propose a novel hierarchi- cal ensemble method able to provide theoretically guaranteed consistent predictions for any Directed Acyclic Graph (DAG)-structured taxonomy, and consequently also for any taxonomy structured according to a tree. Results with a complex real-world DAG-structured taxonomy involving about one thousand classes and twenty thousand of examples show that the proposed hierarchical ensemble approach significantly improves flat methods, especially in terms of precision/recall curves.File | Dimensione | Formato | |
---|---|---|---|
Robinson-valeMCS15-final.pdf
accesso riservato
Tipologia:
Post-print, accepted manuscript ecc. (versione accettata dall'editore)
Dimensione
127.79 kB
Formato
Adobe PDF
|
127.79 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.