Three relevant areas of interest in symbolic Machine Learning are incremental supervised learning, multistrategy learning and predicate invention. In many real-world tasks, new observations may point out the inadequacy of the learned model. In such a case, incremental approaches allow to adjust it, instead of learning a new model from scratch. Specifically, when a negative example is wrongly classified by a model, emph{specialization} refinement operators are needed. A powerful way to specialize a theory in Inductive Logic Programming is adding negated preconditions to concept definitions. This paper describes an empowered specialization operator that allows to introduce the negation of conjunctions of preconditions using predicate invention. An implementation of the operator is proposed, and experiments purposely devised to stress it prove that the proposed approach is correct and viable even under quite complex conditions.

Predicate Invention-based Specialization in Inductive Logic Programming

FERILLI, Stefano
2016-01-01

Abstract

Three relevant areas of interest in symbolic Machine Learning are incremental supervised learning, multistrategy learning and predicate invention. In many real-world tasks, new observations may point out the inadequacy of the learned model. In such a case, incremental approaches allow to adjust it, instead of learning a new model from scratch. Specifically, when a negative example is wrongly classified by a model, emph{specialization} refinement operators are needed. A powerful way to specialize a theory in Inductive Logic Programming is adding negated preconditions to concept definitions. This paper describes an empowered specialization operator that allows to introduce the negation of conjunctions of preconditions using predicate invention. An implementation of the operator is proposed, and experiments purposely devised to stress it prove that the proposed approach is correct and viable even under quite complex conditions.
File in questo prodotto:
File Dimensione Formato  
Ferilli2016_Article_PredicateInvention-basedSpecia.pdf

non disponibili

Tipologia: Documento in Versione Editoriale
Licenza: NON PUBBLICO - Accesso privato/ristretto
Dimensione 759.09 kB
Formato Adobe PDF
759.09 kB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11586/172355
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? 2
social impact