In this work we create a bridge between Convolutional Neural Networks and Answer Set Programming in order to tackle the known Picasso Problem in the automated detection of images. The basic idea is to first exploit the main features of the neural network approach for image recognition, and then to address the problem of identifying well-formed (not meshed up) images by means of explicit knowledge expressed by logical rules. Preliminary experiments suggest that the proposed approach is promising and can be considered as a first step in the direction of solving the Picasso Problem, as well as a witness of the benefits that can be obtained by the combination of a neural approach with a pure symbolic one.

Combining neural and symbolic approaches to solve the Picasso problem: A first step

Gliozzi V.
Co-first
;
Pozzato G. L.
Co-first
;
Valese A.
Co-first
2022-01-01

Abstract

In this work we create a bridge between Convolutional Neural Networks and Answer Set Programming in order to tackle the known Picasso Problem in the automated detection of images. The basic idea is to first exploit the main features of the neural network approach for image recognition, and then to address the problem of identifying well-formed (not meshed up) images by means of explicit knowledge expressed by logical rules. Preliminary experiments suggest that the proposed approach is promising and can be considered as a first step in the direction of solving the Picasso Problem, as well as a witness of the benefits that can be obtained by the combination of a neural approach with a pure symbolic one.
2022
74
102203
102210
https://www.sciencedirect.com/science/article/abs/pii/S0141938222000439
Answer Set Programming; Automated reasoning; Neural networks
Gliozzi V.; Pozzato G.L.; Valese A.
File in questo prodotto:
File Dimensione Formato  
display-preproof-1-s2.0-S0141938222000439-main.pdf

Accesso riservato

Tipo di file: PREPRINT (PRIMA BOZZA)
Dimensione 5.76 MB
Formato Adobe PDF
5.76 MB 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/2318/1859465
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? 1
social impact