We propose a novel method to inject symbolic knowledge in form of Datalog formulæ into neural networks (NN), called KINS (Knowledge Injection via Network Structuring). The idea behind our method is to extend NN internal structure with ad-hoc layers built out the injected symbolic knowledge. KINS does not constrain NN to any specific architecture, neither requires logic formulæ to be ground. Moreover, it is robust w.r.t. both lack of data and imperfect/incomplete knowledge. Experiments are reported to demonstrate the potential of KINS.

Matteo Magnini, G.C. (2022). KINS: Knowledge Injection via Network Structuring. Aachen : Sun SITE Central Europe (CEUR), Technical University of Aachen (RWTH).

KINS: Knowledge Injection via Network Structuring

Matteo Magnini;Giovanni Ciatto;Andrea Omicini
2022

Abstract

We propose a novel method to inject symbolic knowledge in form of Datalog formulæ into neural networks (NN), called KINS (Knowledge Injection via Network Structuring). The idea behind our method is to extend NN internal structure with ad-hoc layers built out the injected symbolic knowledge. KINS does not constrain NN to any specific architecture, neither requires logic formulæ to be ground. Moreover, it is robust w.r.t. both lack of data and imperfect/incomplete knowledge. Experiments are reported to demonstrate the potential of KINS.
2022
Proceedings of the 37th Italian Conference on Computational Logic
254
267
Matteo Magnini, G.C. (2022). KINS: Knowledge Injection via Network Structuring. Aachen : Sun SITE Central Europe (CEUR), Technical University of Aachen (RWTH).
Matteo Magnini, Giovanni Ciatto, Andrea Omicini
File in questo prodotto:
File Dimensione Formato  
paper_25.pdf

accesso aperto

Descrizione: versione finale
Tipo: Versione (PDF) editoriale
Licenza: Licenza per Accesso Aperto. Creative Commons Attribuzione (CCBY)
Dimensione 966.49 kB
Formato Adobe PDF
966.49 kB Adobe PDF Visualizza/Apri

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/11585/899494
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 4
  • ???jsp.display-item.citation.isi??? ND
social impact